Rnnpool facedetection (#215)

* add m4 model

* add rnnpool sparsity

* revert to previous+remove basenet

* scut training and testing

* augmentations and data file changes

* update readme and eval files

* evaluation code

* fix bugs

* remove lists

* data prep script

* merge face detection and m4

* eval arch options

* finetune

* readme changes

* readme update

* readme update

* newlines and slashes

* add dataset directory as environment variable

* rpool face c detect bug

* support for multigpu

* multigpu fixes

* remove subset option

* readme edit

* mkdir

* readme changes

* fix warning

* rnnpool device

* add arch

* trace generation

* Update eval.py

* Update prior_box.py

* Update multibox_loss.py

* Update train.py

* Update scut_test.py

* eval bug + newlines

* Remove stray newline

Co-authored-by: Harsha Vardhan Simhadri <harsha-simhadri@users.noreply.github.com>
Co-authored-by: ShikharJ <jaiswalshikhar87@gmail.com>

examples/pytorch/vision/Face_Detection/README.md

@@ -1,9 +1,10 @@
-# Code for Face Detection experiments with RNNPool
+# Code for Face Detection Experiments with RNNPool
+Refer to README_M4.md for instructions related to the M4 model
 ## Requirements
-1. Follow instructions to install requirements for EdgeML operators and the EdgeML operators [here](https://github.com/microsoft/EdgeML/blob/master/pytorch/README.md).
+1. Follow instructions to install EdgeML operators and their pre-requisites [here](https://github.com/microsoft/EdgeML/blob/master/pytorch/README.md).
 2. Install requirements for face detection model using
 ``` pip install -r requirements.txt ``` 
-We have tested the installation and the code on Ubuntu 18.04 with Cuda 10.2 and CuDNN 7.6
+We have tested the installation and the code on Ubuntu 18.04 with Python 3.6, Cuda 10.2 and CuDNN 7.6
 
 ## Dataset
 1. Download WIDER face dataset images and annotations from http://shuoyang1213.me/WIDERFACE/ and place them all in a folder with name 'WIDER_FACE'. That is, download WIDER_train.zip, WIDER_test.zip, WIDER_val.zip, wider_face_split.zip and place it in WIDER_FACE folder, and unzip files using: 
@@ -18,12 +19,17 @@ cd ..
 
 ```
 
-2. In `data/config.py` , set _C.HOME to the parent directory of the above folder, and set the _C.FACE.WIDER_DIR to the folder path. 
-That is, if the WIDER_FACE folder is created in /mnt folder, then _C.HOME='/mnt'
-_C.FACE.WIDER_DIR='/mnt/WIDER_FACE'.
-Similarly, change `data/config_qvga.py` to set _C.HOME and _C.FACE.WIDER_DIR.
+2. Set environment variable DATA_HOME to the parent directory of the above folder
+That is, if the WIDER_FACE folder is created in /mnt folder
+
+``` export DATA_HOME='/mnt' ```
+
+Note that for Windows '/' should be replaced by '\'.
+For all following commands the environment variable IS_QVGA_MONO has to be set as 0 for using config.py (to use RGB 640x480 images) and as 1 for using config_qvga.py (to use monochrome 320x240 images) as the configuration file.
+
+
 3. Run
-``` python prepare_wider_data.py ```
+``` IS_QVGA_MONO=1 python prepare_wider_data.py ```
 
 
 # Usage
@@ -64,15 +70,15 @@ There are two modes of testing the trained model -- the evaluation mode to gener
 
 #### Evaluation Mode
 
-Given a set of images in <your_image_folder>, `eval/py` generates bounding boxes around faces (where the confidence is higher than certain threshold) and write the images in <your_save_folder>. To evaluate the `rpool_face_best_state.pth` model (stored in ./weights), execute the following command: 
+Given a set of images in <your_image_folder>, `eval/py` generates bounding boxes around faces (where the confidence is higher than certain threshold) and write the images in <your_save_folder>. Specify if the model was trained in multigpu setting in --multigpu. To evaluate the `rpool_face_best_state.pth` model (stored in ./weights), execute the following command: 
 
 ```shell
-IS_QVGA_MONO=0 python eval.py --model_arch RPool_Face_Quant --model ./weights/RPool_Face_Quant_best_state.pth --image_folder <your_image_folder> --save_dir <your_save_folder>
+IS_QVGA_MONO=0 python eval.py --model_arch RPool_Face_Quant --model ./weights/RPool_Face_Quant_best_state.pth --image_folder <your_image_folder> --save_dir <your_save_folder> --multigpu True
 ```
 
 For QVGA:
 ```shell
-IS_QVGA_MONO=1 python eval.py --model_arch RPool_Face_QVGA_monochrome --model ./weights/RPool_Face_QVGA_monochrome_best_state.pth --image_folder <your_image_folder> --save_dir <your_save_folder>
+IS_QVGA_MONO=1 python eval.py --model_arch RPool_Face_QVGA_monochrome --model ./weights/RPool_Face_QVGA_monochrome_best_state.pth --image_folder <your_image_folder> --save_dir <your_save_folder> --multigpu True
 ```
 
 This will save images in <your_save_folder> with bounding boxes around faces, where the confidence is high. Here is an example image with a single bounding box.
@@ -86,15 +92,15 @@ If IS_QVGA_MONO=1 the evaluation code accepts an image of any size and resizes a
 #### WIDER Set Test
 In this mode, we test the generated model against the provided WIDER_FACE validation and test dataset. 
 
-For this, first run the following to generate predictions of the model and store output in the '--save_folder' folder. 
+For this, first run the following to generate predictions of the model and store output in the '--save_folder' folder. Specify if the model was trained in multigpu setting in --multigpu.
 
 ```shell
-IS_QVGA_MONO=0 python wider_test.py --model_arch RPool_Face_Quant --model ./weights/RPool_Face_Quant_best_state.pth --save_folder rpool_face_quant_val --subset val
+IS_QVGA_MONO=0 python wider_test.py --model_arch RPool_Face_Quant --model ./weights/RPool_Face_Quant_best_state.pth --save_folder rpool_face_quant_val --subset val --multigpu True
 ```
 
 For QVGA:
 ```shell
-IS_QVGA_MONO=1 python wider_test.py --model_arch RPool_Face_QVGA_monochrome --model ./weights/RPool_Face_QVGA_monochrome_best_state.pth --save_folder rpool_face_qvgamono_val --subset val
+IS_QVGA_MONO=1 python wider_test.py --model_arch RPool_Face_QVGA_monochrome --model ./weights/RPool_Face_QVGA_monochrome_best_state.pth --save_folder rpool_face_qvgamono_val --subset val --multigpu True
 ```
 
 The above command generates predictions for each image in the "validation" dataset. For each image, a separate prediction file is provided (image_name.txt file in appropriate folder). The first line of the prediction file contains the total number of boxes identified. 
@@ -104,8 +110,8 @@ If IS_QVGA_MONO=1 then testing is done by converting images to monochrome and QV
 
 The architecture RPool_Face_QVGA_monochrome is for QVGA monochrome format while RPool_Face_C and RPool_Face_Quant are for VGA RGB format.
 
-###### For calculating MAP scores:
-Now using these boxes, we can compute the standard MAP score that is widely used in this literature (see [here](https://medium.com/@jonathan_hui/map-mean-average-precision-for-object-detection-45c121a31173) for more details) as follows:
+###### For calculating mAP scores:
+Now using these boxes, we can compute the standard mAP score that is widely used in this literature (see [here](https://medium.com/@jonathan_hui/map-mean-average-precision-for-object-detection-45c121a31173) for more details) as follows:
 
 1. Download eval_tools.zip from http://shuoyang1213.me/WIDERFACE/support/eval_script/eval_tools.zip and unzip in a folder of same name in this directory.
 
@@ -116,7 +122,7 @@ wget http://shuoyang1213.me/WIDERFACE/support/eval_script/eval_tools.zip
 unzip eval_tools.zip
 ```
 
-2. Set up scripts to use the Matlab '.mat' data files in eval_tools/ground_truth folder for MAP calculation: The following installs python files that provide the same functionality as the '.m' matlab scripts in eval_tools folder.
+2. Set up scripts to use the Matlab '.mat' data files in eval_tools/ground_truth folder for mAP calculation: The following installs python files that provide the same functionality as the '.m' matlab scripts in eval_tools folder.
 ``` 
 cd eval_tools
 git clone https://github.com/wondervictor/WiderFace-Evaluation.git
@@ -126,20 +132,20 @@ python3 setup.py build_ext --inplace
 
 3. Run ```python3 evaluation.py -p <your_save_folder> -g <groud truth dir>``` in WiderFace-Evaluation folder
 
-where `prediction_dir` is the '--save_folder' used for `wider_test.py` above and <groud truth dir> is the subfolder `eval_tools/ground_truth`. That is in, WiderFace-Evaluation directory, run: 
+where `-p` is the '--save_folder' used for `wider_test.py` above and <groud truth dir> is the subfolder `eval_tools/ground_truth`. That is in, WiderFace-Evaluation directory, run: 
 
 ```shell
-python3 evaluation.py -p <your_save_folder> -g ../ground_truth
+python3 evaluation.py -p ../../rpool_face_qvgamono_val -g ../ground_truth
 ```
-This script should output the MAP for the WIDER-easy, WIDER-medium, and WIDER-hard subsets of the dataset. Our best performance using RPool_Face_Quant model is: 0.80 (WIDER-easy), 0.78 (WIDER-medium), 0.53 (WIDER-hard). 
+This script should output the mAP for the WIDER-easy, WIDER-medium, and WIDER-hard subsets of the dataset. Our best performance using RPool_Face_Quant model is: 0.80 (WIDER-easy), 0.78 (WIDER-medium), 0.53 (WIDER-hard). 
 
 
 ##### Dump RNNPool Input Output Traces and Weights
 
-To save model weights and/or input output pairs for each patch through RNNPool in numpy format use the command below. Put images which you want to save traces for in <your_image_folder> . Specify output folder for saving model weights in numpy format in <your_save_model_numpy_folder>. Specify output folder for saving input output traces of RNNPool in numpy format in <your_save_traces_numpy_folder>. Note that input traces will be saved in a folder named 'inputs' and output traces in a folder named 'outputs' inside <your_save_traces_numpy_folder>.
+For saving model weights and/or input output pairs for each patch through RNNPool in numpy format use the command below. Put images which you want to save traces for in <your_image_folder> . Specify output folder for saving model weights in numpy format in <your_save_model_numpy_folder>. Specify output folder for saving input output traces of RNNPool in numpy format in <your_save_traces_numpy_folder>. Note that input traces will be saved in a folder named 'inputs' and output traces in a folder named 'outputs' inside <your_save_traces_numpy_folder>.
 
 ```shell
-python3 dump_model.py --model ./weights/RPool_Face_QVGA_monochrome_best_state.pth --model_arch RPool_Face_Quant --image_folder <your_image_folder> --save_model_npy_dir <your_save_model_numpy_folder> --save_traces_npy_dir <your_save_traces_numpy_folder>
+python3 dump_model.py --model ./weights/RPool_Face_QVGA_monochrome_best_state.pth --model_arch RPool_Face_QVGA_monochrome --image_folder <your_image_folder> --save_model_npy_dir <your_save_model_numpy_folder> --save_traces_npy_dir <your_save_traces_numpy_folder>
 ```
 If you wish to save only model weights, do not specify --save_traces_npy_dir. If you wish to save only traces do not specify --save_model_npy_dir.
 

examples/pytorch/vision/Face_Detection/README_M4.md

@@ -0,0 +1,141 @@
+# Code for Face Detection Experiments with RNNPool
+## Requirements
+1. Follow instructions to install EdgeML operators and their pre-requisites [here](https://github.com/microsoft/EdgeML/blob/master/pytorch/README.md).
+2. Install requirements for face detection model using
+``` pip install -r requirements.txt ``` 
+We have tested the installation and the code on Ubuntu 18.04 with Python 3.6, Cuda 10.2 and CuDNN 7.6
+
+## Dataset - WIDER Face
+1. Download WIDER face dataset images and annotations from http://shuoyang1213.me/WIDERFACE/ and place them all in a folder with name 'WIDER_FACE'. That is, download WIDER_train.zip, WIDER_test.zip, WIDER_val.zip, wider_face_split.zip and place it in WIDER_FACE folder, and unzip files using: 
+
+```shell
+cd WIDER_FACE
+unzip WIDER_train.zip
+unzip WIDER_test.zip
+unzip WIDER_val.zip
+unzip wider_face_split.zip
+cd ..
+
+```
+
+2. Set environment variable DATA_HOME to the parent directory of the above folder
+That is, if the WIDER_FACE folder is created in /mnt folder
+
+``` export DATA_HOME='/mnt' ```
+
+Note that for Windows '/' should be replaced by '\'.
+
+
+3. Run
+``` IS_QVGA_MONO=1 python prepare_wider_data.py ```
+
+## Dataset - SCUT Head B
+Download SCUT Head Part B dataset images and annotations from https://github.com/HCIILAB/SCUT-HEAD-Dataset-Release. Unzipping will create a folder by the name 'SCUT_HEAD_Part_B'. Place this folder in the same parent directory as the WIDER_FACE folder.
+
+
+# Usage
+
+## Training
+
+```shell
+
+IS_QVGA_MONO=1 python train.py --batch_size 128 --model_arch RPool_Face_M4 --cuda True --multigpu True --save_folder weights/ --epochs 300 --save_frequency 5000 
+
+```
+This will save checkpoints after every '--save_frequency' number of iterations in a weight file with 'checkpoint.pth' at the end and weights for the best state in a file with 'best_state.pth' at the end. These will be saved in '--save_folder'. For resuming training from a checkpoint, use '--resume <checkpoint_name>.pth' with the above command. For example, 
+
+
+```shell
+
+IS_QVGA_MONO=1 python train.py --batch_size 128 --model_arch RPool_Face_M4 --cuda True --multigpu True --save_folder weights/ --epochs 300 --save_frequency 5000 --resume <checkpoint_name>.pth
+
+```
+
+If IS_QVGA_MONO is 0 then training input images will be 640x640 and RGB. 
+If IS_QVGA_MONO is 1 then training input images will be 320x320 and converted to monochrome. 
+
+Input images for training models are cropped and reshaped to square to maintain consistency with [S3FD](https://arxiv.org/abs/1708.05237). However testing can be done on any size of images, thus we resize testing input image size to have area equal to VGA (640x480)/QVGA (320x240), so that aspect ratio is not changed.
+
+The architecture RPool_Face_QVGA_monochrome and RPool_Face_M4 is for QVGA monochrome format while RPool_Face_C and RPool_Face_Quant are for VGA RGB format.
+
+## Finetuning
+
+To obtain a model better suited for conference room scenarios we finetune our model on the SCUT Head B dataset. Set --finetune as True and pass the model pretrained on WIDER_FACE in --resume as follows:
+
+```shell
+
+IS_QVGA_MONO=1 python train.py --batch_size 64 --model_arch RPool_Face_M4 --cuda True --multigpu True --save_folder weights/ --epochs 300 --save_frequency 5000 --resume ./weights/RPool_Face_M4_best_state.pth --finetune True
+
+```
+
+
+## Test
+There are two modes of testing the trained model -- the evaluation mode to generate bounding boxes for a set of sample images, and the test mode to compute statistics like mAP scores.
+
+#### Evaluation Mode
+
+Given a set of images in <your_image_folder>, `eval/py` generates bounding boxes around faces (where the confidence is higher than certain threshold - 0.5 in this case) and write the images in <your_save_folder>. Specify if the model was trained in multigpu setting in --multigpu. To evaluate the `rpool_face_best_state.pth` model (stored in ./weights), execute the following command: 
+
+```shell
+IS_QVGA_MONO=1 python eval.py --model_arch RPool_Face_M4 --model ./weights/RPool_Face_M4_best_state.pth --image_folder <your_image_folder> --save_dir <your_save_folder> --thresh 0.5  --multigpu True
+```
+
+This will save images in <your_save_folder> with bounding boxes around faces, where the confidence is high. It is recommended to use the model finetuned on SCUT Head for evaluation.
+
+If IS_QVGA_MONO=0 the evaluation code accepts an image of any size and resizes it to 640x480x3 while preserving original image aspect ratio.
+
+If IS_QVGA_MONO=1 the evaluation code accepts an image of any size and resizes and converts it to monochrome to make it 320x240x1 while preserving original image aspect ratio.
+
+
+#### Saving Full Model Traces
+Setting the flag --save_traces as True will save input output traces in two separate .npy files for each image in <your_image_folder>, given the architecture and trained model. Run:
+
+
+```shell
+IS_QVGA_MONO=1 python eval.py --model_arch RPool_Face_M4 --model ./weights/RPool_Face_M4_best_state.pth --image_folder <your_image_folder> --save_dir <your_save_folder> --thresh 0.5  --multigpu True  --save_traces True
+```
+
+For generating traces on SCUT Head images, set <your_image_folder> as $DATA_HOME/SCUT_HEAD_Part_B/JPEGImages/
+
+#### SCUT Head Validation Set Test
+In this mode, we test the generated model against the provided SCUT Head Part B validation dataset. Use the SCUT Head finetuned model for this step.
+
+For this, first run the following to generate predictions of the model and store output in the '--save_folder' folder. Specify if the model was trained in multigpu setting in --multigpu.
+
+```shell
+IS_QVGA_MONO=1 python scut_test.py --model_arch RPool_Face_M4 --model ./weights/RPool_Face_M4_best_state.pth --save_folder rpool_face_m4_val --multigpu True
+```
+
+The above command generates predictions for each image in the "validation" dataset. For each image, a separate prediction file is provided (image_name.txt file in appropriate folder). The first line of the prediction file contains the total number of boxes identified. 
+Then each line in the file corresponds to an identified box. For each box, five numbers are generated: length of the box, height of the box, x-axis offset, y-axis offset, confidence value for presence of a face in the box. 
+
+If IS_QVGA_MONO=1 then testing is done by converting images to monochrome and QVGA, else if IS_QVGA_MONO=0 then testing is done on VGA RGB images.
+
+###### For calculating mAP scores:
+Now using these boxes, we can compute the standard mAP score that is widely used in this literature (see [here](https://medium.com/@jonathan_hui/map-mean-average-precision-for-object-detection-45c121a31173) for more details).
+
+In the current Face_Detection directory run:
+``` 
+git clone https://github.com/wondervictor/WiderFace-Evaluation.git
+cd WiderFace-Evaluation 
+python3 setup.py build_ext --inplace
+mv ../scut_evaluation.py ./
+```
+
+Run ```IS_QVGA_MONO=1 python3 scut_evaluation.py -p ../rpool_face_m4_val ``` in WiderFace-Evaluation folder.
+
+where `-p` is the '--save_folder' used for `scut_test.py` above. 
+
+This script should output the mAP on SCUT Head Part B Validation set. Our best performance using RPool_Face_M4 model is: 0.61.
+
+
+##### Dump RNNPool Input Output Traces and Weights
+
+For saving model weights and/or input output pairs for each patch through RNNPool in numpy format use the command below. Put images which you want to save traces for in <your_image_folder> . Specify output folder for saving model weights in numpy format in <your_save_model_numpy_folder>. Specify output folder for saving input output traces of RNNPool in numpy format in <your_save_traces_numpy_folder>. Note that input traces will be saved in a folder named 'inputs' and output traces in a folder named 'outputs' inside <your_save_traces_numpy_folder>.
+
+```shell
+python3 dump_model.py --model ./weights/RPool_Face_M4_best_state.pth --model_arch RPool_Face_M4 --image_folder <your_image_folder> --save_model_npy_dir <your_save_model_numpy_folder> --save_traces_npy_dir <your_save_traces_numpy_folder>
+```
+If you wish to save only model weights, do not specify --save_traces_npy_dir. If you wish to save only traces do not specify --save_model_npy_dir.
+
+Code has been built upon https://github.com/yxlijun/S3FD.pytorch

examples/pytorch/vision/Face_Detection/data/choose_config.py

@@ -12,4 +12,4 @@ if IS_QVGA_MONO == '1':
 	name = name + '_qvga'
 
 
-cfg = import_module('data.' + name)
+cfg = import_module('data.' + name)

examples/pytorch/vision/Face_Detection/data/config.py

@@ -54,12 +54,8 @@ _C.NEG_POS_RATIOS = 3
 _C.NUM_CLASSES = 2
 _C.USE_NMS = True
 
-# dataset config
-_C.HOME = '/mnt/'  ## change here ----------
-
 # face config
 _C.FACE = EasyDict()
 _C.FACE.TRAIN_FILE = './data/face_train.txt'
 _C.FACE.VAL_FILE = './data/face_val.txt'
-_C.FACE.WIDER_DIR = '/mnt/WIDER_FACE'  ## change here ---------
 _C.FACE.OVERLAP_THRESH = [0.1, 0.35, 0.5]

examples/pytorch/vision/Face_Detection/data/config_qvga.py

@@ -53,12 +53,8 @@ _C.NEG_POS_RATIOS = 3
 _C.NUM_CLASSES = 2
 _C.USE_NMS = True
 
-# dataset config
-_C.HOME = '/mnt/'
-
 # face config
 _C.FACE = EasyDict()
 _C.FACE.TRAIN_FILE = './data/face_train.txt'
 _C.FACE.VAL_FILE = './data/face_val.txt'
-_C.FACE.WIDER_DIR = '/mnt/WIDER_FACE'
 _C.FACE.OVERLAP_THRESH = [0.1, 0.35, 0.5]

examples/pytorch/vision/Face_Detection/data/face_val_scutB.txt

@@ -0,0 +1,405 @@
+JPEGImages/PartB_00723.jpg 10 97 102 108 138 1 38 135 59 60 1 15 102 46 63 1 3 87 24 31 1 92 74 28 34 1 151 90 29 26 1 381 72 14 21 1 394 60 14 14 1 439 27 27 27 1 514 75 18 25 1 
+JPEGImages/PartB_01013.jpg 57 70 585 148 81 1 73 459 92 120 1 43 442 68 71 1 68 394 64 60 1 40 381 39 49 1 54 353 31 32 1 52 338 24 24 1 110 333 29 33 1 139 335 31 36 1 137 365 34 43 1 176 383 47 52 1 196 417 47 57 1 242 424 56 76 1 332 468 92 117 1 487 522 126 145 1 301 298 39 52 1 212 340 32 34 1 228 330 24 29 1 163 332 26 29 1 266 338 15 19 1 358 357 31 49 1 389 361 26 32 1 398 383 40 47 1 323 359 28 35 1 361 330 22 31 1 399 323 25 31 1 413 343 24 29 1 441 355 34 45 1 460 335 23 26 1 472 392 45 62 1 507 387 37 52 1 515 362 28 29 1 518 343 25 24 1 557 343 21 25 1 574 389 43 57 1 613 422 53 65 1 578 365 29 31 1 581 350 19 23 1 614 338 19 24 1 669 376 34 41 1 644 365 27 30 1 665 351 24 27 1 713 352 25 31 1 709 418 55 69 1 772 433 74 87 1 753 400 35 43 1 733 374 29 35 1 742 352 26 29 1 785 361 26 33 1 839 371 24 29 1 861 377 30 36 1 820 410 35 36 1 886 427 55 73 1 867 411 50 52 1 940 409 55 56 1 950 378 36 40 1 908 365 29 35 1 
+JPEGImages/PartB_02018.jpg 7 1 69 14 37 1 21 65 34 43 1 49 87 26 34 1 134 89 22 29 1 147 81 30 36 1 181 84 43 43 1 246 80 55 66 1 
+JPEGImages/PartB_02215.jpg 87 257 412 80 79 1 202 366 41 45 1 108 382 43 47 1 52 345 34 34 1 12 329 31 41 1 45 324 40 30 1 60 310 22 21 1 91 332 25 25 1 112 329 17 19 1 109 307 18 20 1 140 329 28 29 1 144 314 17 19 1 151 273 21 20 1 171 340 27 32 1 191 330 20 25 1 218 325 17 25 1 248 336 28 35 1 260 330 27 26 1 262 307 18 20 1 303 327 33 33 1 313 316 19 18 1 301 301 17 17 1 336 309 18 19 1 343 305 15 14 1 365 310 18 21 1 384 344 37 35 1 410 341 49 52 1 413 387 75 88 1 478 338 39 42 1 444 317 25 28 1 424 314 21 22 1 416 295 18 21 1 403 301 18 23 1 384 311 23 22 1 386 302 18 20 1 463 305 13 12 1 475 308 17 20 1 504 324 24 23 1 552 315 29 34 1 551 307 19 20 1 540 302 16 20 1 535 271 15 18 1 496 305 8 13 1 588 329 30 36 1 554 377 68 70 1 618 378 57 57 1 564 293 12 16 1 585 305 12 19 1 600 310 23 23 1 610 293 15 17 1 633 298 14 16 1 647 306 16 18 1 642 318 20 22 1 634 336 24 28 1 676 327 25 28 1 676 308 15 20 1 680 300 12 14 1 704 304 12 15 1 716 316 15 21 1 704 337 33 33 1 763 301 21 20 1 789 304 20 26 1 782 333 28 34 1 773 327 20 23 1 818 297 15 16 1 820 311 14 20 1 830 314 19 26 1 862 308 21 27 1 877 292 18 17 1 870 284 14 16 1 893 287 12 14 1 893 303 17 22 1 883 322 30 31 1 894 349 43 45 1 916 318 27 31 1 925 301 20 22 1 918 281 16 19 1 943 298 16 23 1 961 314 21 22 1 967 295 18 25 1 987 308 17 27 1 988 328 36 38 1 961 343 39 46 1 1003 307 21 25 1 1028 308 18 21 1 1051 300 20 25 1 1057 325 24 36 1 
+JPEGImages/PartB_01865.jpg 9 9 82 68 74 1 89 74 40 44 1 183 54 52 59 1 147 116 104 107 1 257 67 43 51 1 364 63 40 48 1 398 49 36 40 1 453 75 52 72 1 491 48 50 59 1 
+JPEGImages/PartB_00136.jpg 31 30 181 21 19 1 81 179 16 18 1 108 181 20 16 1 153 178 17 15 1 190 177 13 15 1 225 175 14 17 1 267 174 16 18 1 303 175 16 21 1 350 176 14 18 1 381 169 13 19 1 60 243 38 32 1 101 230 20 23 1 116 234 23 35 1 150 256 42 35 1 176 229 28 25 1 156 223 21 22 1 144 240 29 24 1 34 227 25 19 1 43 242 25 30 1 215 236 25 31 1 241 251 33 21 1 280 220 19 22 1 308 229 22 20 1 298 242 29 25 1 334 243 29 32 1 340 223 22 22 1 365 240 29 37 1 363 225 19 22 1 400 235 22 24 1 393 218 22 23 1 424 223 15 35 1 
+JPEGImages/PartB_00444.jpg 12 43 80 25 32 1 92 69 25 28 1 131 69 28 30 1 182 65 26 34 1 213 63 23 26 1 111 116 74 69 1 269 54 22 30 1 297 62 26 29 1 329 74 29 36 1 321 70 21 27 1 380 81 30 35 1 424 80 41 44 1 
+JPEGImages/PartB_02367.jpg 23 23 121 19 23 1 53 123 20 22 1 74 158 19 23 1 97 122 19 22 1 117 160 17 21 1 132 127 18 18 1 154 128 18 22 1 163 159 20 23 1 185 128 17 19 1 203 157 21 23 1 224 126 18 21 1 246 154 20 25 1 240 125 14 16 1 261 126 15 21 1 276 124 16 21 1 292 157 23 29 1 291 121 18 22 1 322 125 19 21 1 327 156 19 23 1 357 123 20 24 1 373 162 20 23 1 392 122 22 23 1 411 123 20 24 1 
+JPEGImages/PartB_02117.jpg 41 66 346 79 96 1 41 260 44 44 1 106 244 28 34 1 127 263 35 38 1 145 289 60 74 1 157 240 25 34 1 171 249 24 27 1 194 247 28 32 1 236 256 38 40 1 269 284 52 52 1 284 307 56 60 1 278 359 101 110 1 437 317 62 80 1 405 288 43 52 1 410 271 35 36 1 417 236 25 27 1 383 243 24 28 1 335 243 24 25 1 317 230 22 23 1 297 246 31 32 1 278 246 23 24 1 268 227 18 22 1 480 273 24 24 1 496 267 39 44 1 463 207 26 31 1 478 240 25 25 1 523 238 20 24 1 534 245 27 26 1 567 238 23 27 1 564 268 28 35 1 543 292 44 61 1 612 249 41 54 1 599 253 22 25 1 655 266 31 41 1 661 242 23 30 1 681 260 22 26 1 719 253 20 27 1 739 276 29 44 1 755 253 21 25 1 780 263 31 37 1 836 246 51 61 1 
+JPEGImages/PartB_00708.jpg 4 23 100 184 198 1 384 152 153 196 1 546 79 229 303 1 812 60 210 323 1 
+JPEGImages/PartB_00863.jpg 6 68 238 83 74 1 354 116 52 66 1 540 253 53 62 1 319 261 141 190 1 222 224 94 110 1 734 207 116 260 1 
+JPEGImages/PartB_01417.jpg 32 2 195 39 40 1 42 185 30 33 1 67 146 19 16 1 55 150 12 15 1 94 154 11 16 1 107 155 16 19 1 111 183 34 39 1 68 171 24 20 1 133 155 16 17 1 139 146 13 13 1 162 140 10 10 1 154 154 13 14 1 159 168 26 24 1 191 164 20 17 1 208 159 22 24 1 206 149 12 11 1 236 151 17 18 1 253 149 13 17 1 245 142 8 11 1 260 141 12 13 1 269 149 17 18 1 321 142 15 14 1 249 200 46 64 1 285 207 39 47 1 336 147 41 52 1 372 153 22 29 1 386 144 21 22 1 399 182 36 48 1 395 224 63 68 1 420 165 33 32 1 429 144 18 18 1 411 151 15 19 1 
+JPEGImages/PartB_02191.jpg 24 478 273 22 28 1 36 413 50 50 1 104 410 47 55 1 61 470 86 82 1 172 438 63 54 1 224 395 41 31 1 299 376 30 38 1 336 360 31 35 1 344 393 42 48 1 289 471 93 84 1 406 398 45 47 1 379 387 36 38 1 470 375 33 42 1 505 374 31 30 1 501 419 56 58 1 552 369 33 30 1 578 369 23 31 1 601 367 30 33 1 646 392 43 44 1 650 441 50 50 1 701 434 47 53 1 687 382 29 37 1 691 362 26 27 1 743 365 40 43 1 
+JPEGImages/PartB_01027.jpg 4 273 444 154 176 1 419 307 95 128 1 531 234 76 116 1 632 194 60 81 1 
+JPEGImages/PartB_00946.jpg 52 78 87 10 11 1 14 116 15 14 1 2 122 12 11 1 11 138 13 18 1 24 142 10 14 1 36 142 13 17 1 53 140 10 14 1 62 138 15 13 1 47 119 9 11 1 319 164 66 64 1 402 171 44 47 1 450 157 57 60 1 449 134 28 31 1 487 130 26 28 1 383 136 29 32 1 375 124 26 28 1 320 146 34 33 1 293 139 24 30 1 256 154 37 43 1 238 145 30 35 1 191 156 32 31 1 180 177 44 44 1 135 168 28 30 1 84 156 25 23 1 105 146 19 19 1 129 145 12 13 1 149 144 24 24 1 135 127 15 15 1 119 123 12 14 1 112 118 9 9 1 103 128 14 16 1 84 134 12 11 1 75 130 9 11 1 66 124 10 9 1 131 111 13 14 1 145 116 13 11 1 156 128 16 14 1 170 124 11 12 1 186 133 17 20 1 196 140 22 18 1 194 113 11 12 1 207 123 12 13 1 266 127 22 21 1 246 132 14 12 1 224 129 11 13 1 223 112 12 8 1 216 108 9 10 1 251 114 9 9 1 249 106 10 9 1 239 111 8 10 1 275 118 11 9 1 266 105 9 7 1 
+JPEGImages/PartB_00229.jpg 32 70 130 9 13 1 231 156 11 12 1 263 156 14 16 1 300 155 13 15 1 331 151 12 14 1 371 159 13 13 1 397 183 48 47 1 400 167 23 22 1 425 236 69 77 1 468 314 112 110 1 512 208 36 41 1 488 200 27 31 1 477 180 26 25 1 514 177 20 23 1 541 185 22 27 1 557 231 59 65 1 552 208 47 41 1 570 192 34 37 1 277 184 20 14 1 263 178 17 20 1 250 188 20 23 1 236 173 15 23 1 230 195 20 25 1 213 200 27 26 1 185 205 27 31 1 139 197 29 36 1 161 189 20 26 1 99 218 34 41 1 51 220 55 67 1 91 191 22 22 1 111 180 20 28 1 85 179 22 19 1 
+JPEGImages/PartB_01187.jpg 41 98 606 173 162 1 35 510 112 119 1 145 482 73 75 1 99 467 51 56 1 65 452 43 44 1 23 457 43 42 1 36 427 37 38 1 97 433 38 32 1 139 425 35 41 1 165 437 29 29 1 228 426 28 38 1 253 429 32 40 1 300 445 35 43 1 341 428 30 45 1 383 430 45 55 1 370 482 82 120 1 512 533 154 183 1 485 458 58 67 1 478 431 44 48 1 420 412 33 49 1 370 420 26 35 1 282 422 25 33 1 482 366 24 33 1 538 346 24 34 1 585 442 44 50 1 573 419 30 43 1 638 449 50 66 1 682 457 66 86 1 668 422 37 45 1 648 405 28 40 1 697 406 29 38 1 736 412 33 41 1 732 439 46 65 1 833 462 64 81 1 828 432 47 50 1 782 419 42 40 1 841 399 47 50 1 896 399 42 43 1 990 407 34 43 1 938 455 72 81 1 980 485 44 93 1 
+JPEGImages/PartB_00503.jpg 7 3 256 95 110 1 141 221 43 44 1 210 227 46 47 1 265 205 43 49 1 332 206 45 54 1 478 207 81 84 1 315 127 47 49 1 
+JPEGImages/PartB_01513.jpg 23 49 90 46 56 1 89 86 25 33 1 110 90 22 31 1 121 85 20 30 1 137 92 16 26 1 150 93 15 23 1 168 95 11 18 1 176 94 11 18 1 187 95 15 18 1 198 91 12 16 1 210 96 18 19 1 233 95 14 18 1 231 79 15 15 1 262 79 16 20 1 310 74 14 16 1 319 81 16 19 1 354 5 38 35 1 10 75 27 29 1 13 61 13 15 1 8 21 16 18 1 89 69 12 10 1 138 70 14 17 1 174 75 14 15 1 
+JPEGImages/PartB_00117.jpg 8 76 117 48 54 1 62 161 17 21 1 126 160 26 36 1 153 161 38 44 1 140 203 70 79 1 214 179 56 55 1 252 158 43 53 1 340 195 49 75 1 
+JPEGImages/PartB_02131.jpg 13 16 161 39 44 1 50 168 51 52 1 83 159 27 36 1 86 42 33 37 1 99 179 43 45 1 146 221 75 81 1 193 164 30 34 1 221 174 52 61 1 247 157 37 39 1 285 154 36 42 1 370 205 70 78 1 351 178 44 54 1 372 160 38 43 1 
+JPEGImages/PartB_01535.jpg 16 47 217 72 81 1 175 240 70 76 1 164 167 42 53 1 217 146 34 43 1 247 165 50 56 1 307 120 31 38 1 329 139 36 37 1 380 127 31 38 1 385 194 49 47 1 451 142 43 50 1 460 240 72 89 1 561 270 83 85 1 657 221 67 79 1 633 139 39 48 1 565 135 29 35 1 506 197 46 58 1 
+JPEGImages/PartB_02392.jpg 36 416 194 51 54 1 356 174 30 36 1 311 196 47 46 1 224 202 70 69 1 182 170 23 24 1 130 174 32 31 1 90 172 21 25 1 53 176 36 38 1 29 175 23 26 1 15 162 17 17 1 5 154 16 15 1 1 164 20 23 1 39 153 14 15 1 72 154 13 16 1 99 160 14 13 1 116 160 15 16 1 162 165 17 16 1 205 156 16 15 1 226 168 22 20 1 229 154 16 17 1 258 162 23 23 1 252 155 12 14 1 135 154 10 12 1 298 161 11 15 1 287 153 13 14 1 304 149 8 10 1 312 154 15 15 1 327 159 11 12 1 328 150 9 9 1 344 155 11 11 1 373 164 22 27 1 392 161 18 22 1 406 158 17 21 1 400 149 11 15 1 374 148 12 13 1 449 161 22 24 1 
+JPEGImages/PartB_01310.jpg 5 1 56 41 74 1 30 231 121 103 1 114 36 52 71 1 240 58 61 73 1 267 225 111 133 1 
+JPEGImages/PartB_01671.jpg 4 20 76 101 104 1 138 92 63 86 1 195 97 104 129 1 281 69 95 111 1 
+JPEGImages/PartB_02042.jpg 9 85 342 103 86 1 236 195 19 23 1 278 272 44 55 1 342 293 67 78 1 396 240 32 34 1 495 226 22 26 1 541 241 25 27 1 540 228 10 15 1 548 227 14 17 1 
+JPEGImages/PartB_01926.jpg 28 109 209 36 33 1 170 207 40 39 1 94 193 36 32 1 64 193 23 21 1 159 184 25 27 1 159 166 20 21 1 193 185 23 28 1 223 193 22 30 1 242 198 27 30 1 271 206 32 30 1 265 169 22 26 1 247 174 17 18 1 231 181 16 19 1 214 176 17 17 1 305 164 14 15 1 314 183 14 20 1 330 187 23 27 1 324 166 17 19 1 385 160 20 24 1 410 189 34 37 1 451 193 30 31 1 443 188 20 19 1 436 170 20 22 1 419 160 20 26 1 470 184 16 18 1 491 188 26 29 1 512 178 18 29 1 535 191 15 27 1 
+JPEGImages/PartB_00368.jpg 71 275 144 25 31 1 268 123 19 21 1 275 109 15 17 1 279 98 14 17 1 285 87 10 14 1 284 77 9 8 1 289 72 9 8 1 365 167 35 51 1 345 146 26 30 1 330 125 19 25 1 327 112 19 19 1 314 96 14 18 1 305 90 14 13 1 328 84 12 12 1 338 89 13 18 1 352 98 16 19 1 364 106 17 21 1 379 124 18 27 1 328 75 10 10 1 308 83 9 9 1 36 156 38 41 1 1 152 24 32 1 26 131 21 23 1 35 123 18 16 1 64 131 26 30 1 104 138 28 29 1 152 120 20 25 1 172 108 15 16 1 195 100 12 15 1 207 93 10 13 1 213 89 9 11 1 221 82 8 11 1 232 80 7 9 1 237 78 7 9 1 218 72 7 7 1 214 79 6 10 1 205 82 9 11 1 194 87 9 12 1 183 92 11 11 1 167 98 12 11 1 145 110 13 15 1 110 120 20 17 1 83 116 15 20 1 88 104 15 15 1 108 100 14 16 1 118 107 14 15 1 139 97 11 13 1 134 89 11 13 1 154 89 12 14 1 164 84 10 11 1 176 83 10 13 1 183 76 9 8 1 189 81 9 9 1 142 76 6 9 1 129 76 9 10 1 123 73 8 9 1 116 76 8 6 1 112 80 9 6 1 117 85 7 9 1 104 84 7 10 1 97 85 9 13 1 83 83 9 7 1 80 87 11 10 1 76 92 9 10 1 65 87 10 11 1 54 88 10 12 1 49 99 12 16 1 35 89 12 11 1 19 95 13 13 1 11 104 16 15 1 1 97 12 17 1 
+JPEGImages/PartB_00844.jpg 14 31 120 21 19 1 52 130 22 24 1 71 121 14 16 1 85 139 20 25 1 95 145 26 27 1 142 152 46 40 1 161 121 18 24 1 176 128 18 22 1 212 133 19 26 1 252 134 22 24 1 284 136 21 29 1 318 141 21 29 1 390 149 28 38 1 408 157 57 54 1 
+JPEGImages/PartB_00173.jpg 8 326 278 49 62 1 428 253 53 61 1 512 293 54 70 1 617 294 46 58 1 697 311 49 59 1 781 297 51 65 1 878 302 55 59 1 972 285 52 52 1 
+JPEGImages/PartB_01684.jpg 3 9 25 111 141 1 230 34 172 232 1 478 36 156 177 1 
+JPEGImages/PartB_01019.jpg 7 86 141 18 20 1 119 143 20 24 1 166 157 16 19 1 198 154 19 21 1 317 150 18 21 1 368 159 15 21 1 406 160 15 18 1 
+JPEGImages/PartB_01345.jpg 29 18 365 42 37 1 36 396 40 31 1 64 424 55 52 1 99 461 79 66 1 215 429 65 53 1 173 375 45 52 1 113 375 46 45 1 115 341 37 38 1 164 348 35 36 1 221 382 38 33 1 284 382 43 39 1 306 373 39 32 1 267 347 32 27 1 344 344 35 32 1 418 370 39 38 1 400 391 39 33 1 359 417 63 58 1 355 448 85 85 1 417 342 26 25 1 474 256 19 25 1 498 347 28 33 1 518 367 38 39 1 513 396 47 33 1 522 414 60 62 1 571 345 29 35 1 584 370 36 40 1 633 390 50 45 1 668 410 56 52 1 627 430 100 105 1 
+JPEGImages/PartB_00100.jpg 13 145 101 15 18 1 238 117 14 16 1 293 162 12 14 1 306 170 13 15 1 323 169 13 17 1 332 182 15 19 1 348 185 22 25 1 57 230 39 34 1 71 213 30 30 1 78 194 24 26 1 97 163 15 16 1 90 157 11 10 1 71 183 20 18 1 
+JPEGImages/PartB_00086.jpg 23 165 301 32 37 1 226 427 47 57 1 280 411 61 65 1 288 380 37 40 1 257 370 33 42 1 370 360 42 37 1 353 396 47 49 1 536 464 74 90 1 528 407 46 50 1 492 366 39 39 1 542 382 39 44 1 561 356 30 31 1 592 370 47 39 1 638 396 54 53 1 725 413 78 80 1 723 380 41 41 1 781 385 45 48 1 771 357 38 33 1 807 349 27 33 1 821 379 32 43 1 860 351 32 30 1 903 388 71 74 1 931 346 32 41 1 
+JPEGImages/PartB_01556.jpg 13 92 439 33 40 1 115 429 45 50 1 170 445 69 75 1 221 430 30 45 1 294 449 42 53 1 361 452 52 64 1 343 436 22 34 1 410 448 31 38 1 480 434 27 39 1 518 441 33 46 1 591 444 36 42 1 629 465 49 50 1 915 218 147 155 1 
+JPEGImages/PartB_02045.jpg 13 26 151 27 29 1 27 180 46 48 1 67 172 38 29 1 100 152 20 24 1 246 145 23 28 1 272 154 28 33 1 304 151 20 24 1 348 170 40 50 1 361 160 30 29 1 299 125 17 18 1 398 148 19 23 1 444 162 44 49 1 451 147 19 24 1 
+JPEGImages/PartB_01141.jpg 11 97 139 57 56 1 55 123 31 28 1 56 110 20 20 1 145 106 20 22 1 179 153 61 61 1 223 106 23 23 1 298 120 42 34 1 342 112 28 28 1 370 153 77 69 1 448 133 41 44 1 443 123 25 24 1 
+JPEGImages/PartB_02345.jpg 28 59 131 34 36 1 31 118 30 34 1 2 107 21 24 1 21 97 19 24 1 14 84 15 14 1 54 104 25 25 1 96 94 14 17 1 107 108 28 32 1 136 93 13 15 1 156 87 16 20 1 166 28 59 71 1 228 80 15 20 1 242 97 21 26 1 254 86 15 15 1 268 77 11 16 1 275 87 12 14 1 286 89 10 15 1 302 96 23 25 1 316 114 38 40 1 312 81 15 15 1 326 82 11 16 1 332 84 18 23 1 354 97 20 25 1 362 115 24 32 1 365 60 13 13 1 377 79 13 18 1 398 88 15 19 1 402 76 13 15 1 
+JPEGImages/PartB_02220.jpg 24 149 160 35 43 1 195 172 40 45 1 187 156 20 28 1 139 161 20 26 1 126 160 20 24 1 227 163 29 32 1 248 164 27 33 1 296 159 21 23 1 308 194 50 60 1 330 178 31 34 1 330 164 17 21 1 376 159 23 26 1 392 161 26 32 1 427 187 44 56 1 460 158 24 32 1 487 170 29 37 1 494 165 23 21 1 548 182 35 33 1 585 174 22 23 1 609 167 18 26 1 644 179 36 44 1 669 170 23 29 1 702 195 31 38 1 705 175 25 30 1 
+JPEGImages/PartB_02065.jpg 32 54 139 21 26 1 71 158 27 29 1 98 165 30 31 1 114 144 21 27 1 163 141 19 22 1 146 177 37 43 1 205 193 42 48 1 260 243 53 69 1 365 207 35 36 1 391 184 31 35 1 417 167 27 30 1 446 164 22 24 1 449 155 16 17 1 463 149 14 18 1 456 140 16 18 1 467 135 11 16 1 467 147 20 19 1 442 131 8 12 1 427 138 13 12 1 415 132 10 12 1 398 133 11 14 1 394 128 10 13 1 373 130 13 18 1 353 129 11 13 1 353 123 10 10 1 330 126 12 14 1 314 125 12 15 1 307 127 13 14 1 285 131 10 14 1 261 132 12 15 1 230 131 14 17 1 193 135 18 22 1 
+JPEGImages/PartB_01497.jpg 7 290 366 124 123 1 290 503 127 152 1 308 76 93 106 1 527 275 75 92 1 684 459 139 141 1 738 266 87 104 1 914 335 120 122 1 
+JPEGImages/PartB_01585.jpg 10 75 72 136 120 1 82 30 97 76 1 271 8 93 123 1 307 46 100 108 1 378 95 70 84 1 445 111 98 117 1 621 117 95 118 1 600 53 89 116 1 735 112 63 102 1 778 97 80 109 1 
+JPEGImages/PartB_00522.jpg 37 49 187 37 43 1 50 120 33 37 1 113 81 31 39 1 125 192 41 42 1 201 193 39 35 1 272 200 38 39 1 345 195 33 43 1 417 194 31 38 1 497 181 32 42 1 566 187 40 39 1 636 192 38 47 1 714 196 34 42 1 691 129 30 40 1 625 117 30 35 1 566 126 33 33 1 497 133 37 45 1 425 131 26 35 1 366 143 33 44 1 304 134 36 35 1 228 134 43 42 1 157 136 31 41 1 167 80 33 40 1 239 80 30 35 1 278 76 30 39 1 323 70 33 34 1 377 76 30 30 1 428 70 34 33 1 485 74 32 37 1 541 77 31 37 1 567 47 33 31 1 621 26 35 35 1 451 26 31 34 1 395 36 29 41 1 328 34 29 34 1 245 26 33 35 1 194 28 33 34 1 154 28 28 33 1 
+JPEGImages/PartB_01879.jpg 10 15 165 53 56 1 92 155 46 54 1 157 165 46 53 1 216 186 42 44 1 265 179 39 49 1 310 182 37 44 1 378 155 49 53 1 426 133 53 65 1 511 153 54 59 1 570 149 59 71 1 
+JPEGImages/PartB_01455.jpg 44 156 307 83 93 1 83 282 64 64 1 56 252 48 44 1 62 209 43 46 1 9 217 45 39 1 146 208 31 35 1 179 211 28 34 1 199 220 34 40 1 218 203 27 23 1 246 207 50 51 1 274 207 36 33 1 289 250 54 58 1 343 254 57 64 1 368 236 34 38 1 337 222 31 33 1 348 207 28 28 1 330 190 23 25 1 322 181 18 28 1 279 184 23 22 1 368 183 24 26 1 409 183 28 28 1 421 203 28 32 1 428 225 34 43 1 469 235 43 59 1 499 222 34 40 1 479 203 28 33 1 516 180 25 29 1 536 189 27 31 1 562 205 38 37 1 583 180 26 32 1 607 188 24 32 1 641 177 27 33 1 598 414 121 121 1 683 312 75 99 1 753 264 69 74 1 736 231 49 49 1 759 209 38 43 1 792 198 30 37 1 806 184 32 33 1 913 258 37 74 1 896 249 43 56 1 900 211 39 46 1 919 197 30 37 1 910 182 32 35 1 
+JPEGImages/PartB_01211.jpg 24 66 1083 168 161 1 103 994 102 103 1 259 1079 164 184 1 443 1060 125 141 1 414 1033 88 91 1 280 1476 297 252 1 761 1090 141 180 1 820 969 112 123 1 957 1088 152 154 1 1046 992 124 125 1 1012 1272 208 243 1 1323 683 113 155 1 1482 761 84 134 1 1323 972 97 116 1 1486 1045 134 136 1 1605 936 127 129 1 1600 1236 189 233 1 2059 1435 327 293 1 1873 1042 134 164 1 1886 949 142 123 1 2007 995 114 111 1 2127 1058 202 177 1 2314 1013 132 156 1 2321 965 86 100 1 
+JPEGImages/PartB_01042.jpg 14 1 224 30 39 1 56 220 28 24 1 93 214 21 22 1 134 251 28 26 1 92 237 25 30 1 202 234 34 37 1 224 229 20 27 1 229 217 16 27 1 243 216 17 30 1 308 153 23 23 1 433 217 26 30 1 472 231 34 36 1 456 233 23 33 1 533 204 45 55 1 
+JPEGImages/PartB_01035.jpg 7 11 372 105 137 1 114 441 56 67 1 187 358 46 72 1 396 448 51 60 1 529 424 72 77 1 593 464 150 141 1 837 386 106 136 1 
+JPEGImages/PartB_01937.jpg 12 24 280 36 43 1 2 288 31 67 1 87 285 61 71 1 126 282 35 50 1 220 290 34 46 1 263 280 34 42 1 335 285 48 50 1 352 274 22 27 1 561 318 97 105 1 656 306 123 144 1 841 214 42 45 1 867 440 156 175 1 
+JPEGImages/PartB_01795.jpg 6 267 101 46 57 1 315 120 34 42 1 326 65 64 79 1 376 99 34 47 1 397 71 95 94 1 489 75 50 79 1 
+JPEGImages/PartB_02021.jpg 17 2 191 38 46 1 41 188 30 32 1 54 206 48 58 1 85 192 24 24 1 101 186 36 42 1 103 175 25 25 1 130 189 29 35 1 140 180 25 24 1 163 178 22 25 1 156 120 15 21 1 159 206 84 86 1 200 178 40 43 1 290 186 36 42 1 298 175 25 26 1 334 220 66 63 1 351 196 25 22 1 380 194 19 31 1 
+JPEGImages/PartB_01724.jpg 20 87 67 40 40 1 273 142 19 20 1 287 140 21 23 1 304 152 24 26 1 324 168 24 30 1 352 178 32 33 1 383 199 37 38 1 364 134 17 19 1 395 135 18 21 1 407 147 16 26 1 417 148 24 30 1 428 126 16 18 1 447 129 18 25 1 472 167 30 30 1 469 229 55 56 1 514 193 36 41 1 499 138 24 25 1 539 150 25 25 1 580 158 19 29 1 580 137 20 22 1 
+JPEGImages/PartB_01813.jpg 4 92 228 59 52 1 176 247 65 53 1 246 88 27 34 1 355 226 45 58 1 
+JPEGImages/PartB_00694.jpg 3 35 230 147 168 1 319 237 201 233 1 683 117 287 355 1 
+JPEGImages/PartB_00041.jpg 18 539 450 68 72 1 293 636 112 120 1 148 761 189 205 1 161 641 125 134 1 11 645 100 102 1 1 731 79 117 1 720 689 157 163 1 861 702 207 218 1 828 613 113 112 1 1175 656 137 148 1 1266 675 111 166 1 1352 532 84 95 1 1100 657 107 131 1 1911 716 166 207 1 2071 689 120 134 1 1870 673 132 145 1 2387 723 156 172 1 2300 647 132 187 1 
+JPEGImages/PartB_01806.jpg 14 2 129 27 32 1 1 97 12 15 1 22 98 16 17 1 35 89 13 14 1 61 100 16 18 1 61 91 15 13 1 45 139 36 37 1 104 111 21 21 1 96 98 15 17 1 138 106 16 18 1 172 123 25 28 1 202 101 20 24 1 303 95 29 31 1 332 49 52 57 1 
+JPEGImages/PartB_00535.jpg 8 49 106 95 109 1 76 1 57 52 1 161 3 64 72 1 264 87 86 86 1 323 180 123 132 1 385 131 108 117 1 302 5 62 71 1 457 24 68 74 1 
+JPEGImages/PartB_02277.jpg 4 3 86 83 204 1 78 190 68 79 1 124 211 78 99 1 362 175 145 184 1 
+JPEGImages/PartB_00255.jpg 18 24 147 32 42 1 66 149 24 30 1 101 150 31 36 1 142 142 40 53 1 212 149 24 32 1 194 162 14 23 1 250 157 36 40 1 301 145 30 44 1 337 152 25 42 1 379 144 33 39 1 418 154 30 38 1 448 152 23 26 1 473 156 15 23 1 492 164 21 30 1 513 171 19 23 1 523 158 14 17 1 546 166 22 31 1 576 166 21 20 1 
+JPEGImages/PartB_00221.jpg 18 1761 1231 96 120 1 5145 1285 104 127 1 4026 1366 58 88 1 4745 2489 523 558 1 3661 2285 350 412 1 4026 1943 215 246 1 4076 1801 185 211 1 3518 1927 189 208 1 3257 2212 315 304 1 3203 1874 127 173 1 2699 2024 231 234 1 2641 1889 189 177 1 2391 1885 208 242 1 1249 3012 658 569 1 1295 2454 473 462 1 1311 2027 257 300 1 288 2039 203 223 1 314 1966 162 146 1 
+JPEGImages/PartB_01032.jpg 19 245 153 15 14 1 317 202 24 28 1 389 217 35 37 1 417 193 33 37 1 266 210 25 24 1 231 222 36 49 1 146 265 78 65 1 168 229 32 36 1 212 209 21 24 1 170 207 20 23 1 111 204 20 24 1 87 228 28 29 1 51 217 26 26 1 73 208 14 16 1 42 207 16 16 1 25 204 15 15 1 17 220 27 25 1 1 265 50 58 1 1 214 10 20 1 
+JPEGImages/PartB_01818.jpg 3 160 133 74 85 1 466 307 87 74 1 624 287 76 72 1 
+JPEGImages/PartB_01906.jpg 51 349 180 67 77 1 425 143 26 38 1 439 133 23 26 1 453 127 16 19 1 466 125 15 18 1 440 122 11 14 1 433 131 7 12 1 425 134 10 12 1 416 134 12 13 1 405 139 19 18 1 413 125 13 14 1 421 122 10 12 1 392 129 14 12 1 376 139 22 21 1 359 137 18 19 1 311 140 23 25 1 192 166 43 46 1 260 138 18 15 1 284 108 8 9 1 331 127 10 13 1 344 124 10 11 1 318 130 9 9 1 295 130 14 14 1 272 131 9 10 1 275 135 10 8 1 287 130 9 12 1 248 135 15 16 1 246 130 10 11 1 234 130 11 14 1 219 133 17 18 1 224 128 12 10 1 205 142 19 21 1 207 136 12 11 1 192 133 16 18 1 176 134 11 14 1 186 151 17 14 1 151 139 18 17 1 165 132 10 12 1 184 128 12 15 1 140 134 13 10 1 135 146 23 17 1 126 134 12 15 1 100 143 21 19 1 109 139 13 13 1 77 137 14 14 1 64 159 21 21 1 34 154 30 25 1 46 142 22 19 1 67 136 12 14 1 2 165 35 36 1 3 148 15 14 1 
+JPEGImages/PartB_00244.jpg 17 32 195 62 58 1 19 156 43 41 1 31 128 36 38 1 127 130 33 35 1 166 146 39 39 1 174 161 41 43 1 244 194 56 47 1 248 131 34 35 1 272 115 26 27 1 335 146 33 34 1 232 300 83 88 1 376 212 51 63 1 398 155 38 42 1 513 173 43 51 1 528 249 63 79 1 542 88 22 23 1 717 170 51 47 1 
+JPEGImages/PartB_00559.jpg 31 156 177 137 161 1 139 98 88 96 1 124 54 36 42 1 159 45 32 38 1 137 39 26 21 1 139 28 20 20 1 218 29 28 33 1 246 40 30 38 1 266 60 43 46 1 287 41 27 27 1 321 31 27 36 1 333 77 52 61 1 379 104 62 76 1 518 173 103 133 1 547 101 67 84 1 453 73 42 52 1 386 64 38 42 1 362 41 27 28 1 403 31 28 43 1 442 43 32 31 1 481 38 21 26 1 502 44 25 31 1 523 37 24 35 1 557 49 27 34 1 586 79 39 52 1 611 55 38 51 1 695 70 29 36 1 739 109 56 63 1 731 160 96 109 1 739 74 51 65 1 833 85 44 58 1 
+JPEGImages/PartB_01177.jpg 13 37 687 149 136 1 118 566 110 110 1 35 505 101 104 1 269 526 101 104 1 387 576 107 124 1 452 499 88 89 1 603 508 88 101 1 687 485 67 78 1 623 282 83 82 1 949 690 188 157 1 1068 616 117 150 1 1081 530 101 122 1 1105 494 101 105 1 
+JPEGImages/PartB_00670.jpg 6 97 189 78 107 1 161 1 75 89 1 187 205 168 206 1 462 191 154 196 1 634 38 101 118 1 533 157 76 90 1 
+JPEGImages/PartB_00713.jpg 24 49 449 66 68 1 75 417 52 48 1 103 374 43 51 1 203 355 30 29 1 228 390 47 41 1 187 414 64 70 1 260 427 61 69 1 279 438 84 82 1 284 361 32 29 1 338 385 44 45 1 348 424 42 53 1 393 412 58 63 1 393 385 36 39 1 430 391 41 40 1 439 462 87 104 1 505 394 53 63 1 556 496 109 104 1 579 360 38 48 1 618 384 36 38 1 672 387 40 41 1 713 384 31 43 1 730 364 33 38 1 740 395 51 57 1 688 291 24 25 1 
+JPEGImages/PartB_01014.jpg 11 16 179 257 207 1 42 5 73 67 1 131 117 87 84 1 250 108 118 127 1 384 110 173 183 1 510 71 117 115 1 365 26 85 84 1 756 13 135 162 1 852 110 252 282 1 1032 17 168 198 1 525 3 52 37 1 
+JPEGImages/PartB_01143.jpg 35 37 211 50 53 1 14 197 31 33 1 1 174 27 33 1 97 246 62 66 1 225 248 57 57 1 148 211 43 48 1 113 201 35 40 1 82 178 38 37 1 184 187 28 28 1 199 205 41 31 1 249 220 42 40 1 238 190 28 33 1 285 208 31 34 1 333 245 49 46 1 425 249 47 46 1 351 224 30 35 1 315 193 22 32 1 352 203 27 28 1 369 215 24 25 1 412 215 27 29 1 410 197 28 27 1 433 219 39 37 1 465 214 22 25 1 470 203 17 25 1 494 224 26 32 1 517 245 41 42 1 522 221 22 29 1 515 202 22 28 1 557 194 29 31 1 561 218 34 36 1 591 240 43 55 1 644 254 34 46 1 659 235 26 37 1 627 218 33 39 1 604 217 26 30 1 
+JPEGImages/PartB_00630.jpg 23 220 258 142 157 1 84 167 121 128 1 1 153 72 97 1 187 121 52 52 1 238 102 42 53 1 261 131 37 48 1 289 89 39 46 1 322 115 36 42 1 360 115 48 56 1 418 115 47 58 1 417 39 31 50 1 472 98 32 43 1 491 112 38 36 1 512 133 47 65 1 524 94 29 35 1 553 103 36 43 1 609 131 46 62 1 638 110 29 32 1 673 99 30 42 1 710 101 30 39 1 775 112 35 50 1 836 110 33 45 1 962 102 50 62 1 
+JPEGImages/PartB_01307.jpg 10 1148 1463 325 357 1 966 1935 336 378 1 195 1988 318 336 1 1627 1952 314 386 1 2016 1702 361 418 1 2398 2135 250 275 1 2795 1817 393 400 1 3391 1345 432 511 1 4152 1820 332 379 1 4534 924 432 490 1 
+JPEGImages/PartB_01630.jpg 29 338 164 19 18 1 332 157 15 18 1 320 158 11 12 1 293 155 13 14 1 282 149 9 9 1 252 151 13 13 1 244 168 32 33 1 206 151 12 11 1 228 142 7 8 1 194 140 8 8 1 171 141 11 11 1 163 134 6 6 1 130 113 8 11 1 143 137 12 11 1 138 131 9 8 1 122 131 13 16 1 92 157 31 37 1 95 143 21 21 1 86 128 11 13 1 64 127 11 13 1 33 124 11 12 1 48 127 13 13 1 52 119 8 9 1 77 123 7 10 1 101 127 7 6 1 26 118 8 9 1 14 121 9 9 1 12 117 8 9 1 161 163 25 26 1 
+JPEGImages/PartB_00001.jpg 9 45 113 26 23 1 30 106 20 24 1 142 152 101 125 1 192 101 37 35 1 276 142 51 63 1 323 138 58 65 1 296 103 23 23 1 412 124 53 52 1 601 105 28 30 1 
+JPEGImages/PartB_02039.jpg 11 5 47 24 24 1 31 61 23 23 1 74 57 19 21 1 113 79 28 34 1 120 52 20 23 1 162 56 24 24 1 194 79 28 34 1 204 55 17 24 1 243 52 21 27 1 281 59 23 25 1 326 42 24 25 1 
+JPEGImages/PartB_00975.jpg 9 260 163 121 146 1 408 167 120 166 1 264 349 178 159 1 602 349 131 177 1 592 155 121 195 1 750 187 127 155 1 855 326 134 179 1 1012 341 142 167 1 982 186 126 152 1 
+JPEGImages/PartB_01559.jpg 12 10 319 93 91 1 155 312 58 55 1 244 330 57 62 1 298 312 67 74 1 429 321 41 48 1 491 298 121 130 1 662 255 39 41 1 596 331 29 43 1 696 307 64 81 1 661 330 34 42 1 780 238 135 167 1 4 174 49 95 1 
+JPEGImages/PartB_01707.jpg 21 64 153 72 78 1 276 186 12 16 1 304 185 13 20 1 315 177 11 13 1 326 189 15 19 1 328 175 13 15 1 352 170 12 12 1 366 172 12 17 1 355 188 14 17 1 380 184 18 22 1 376 164 12 14 1 393 176 15 16 1 399 162 12 15 1 428 167 15 17 1 420 187 19 24 1 457 188 20 23 1 478 167 19 22 1 473 155 12 15 1 511 152 15 20 1 543 163 25 28 1 509 190 28 27 1 
+JPEGImages/PartB_01318.jpg 5 59 360 46 47 1 158 169 33 38 1 380 148 47 59 1 342 338 60 69 1 456 132 56 76 1 
+JPEGImages/PartB_02292.jpg 6 56 104 103 111 1 230 124 72 83 1 306 66 79 103 1 417 95 41 46 1 501 78 36 38 1 533 71 36 40 1 
+JPEGImages/PartB_01557.jpg 8 10 376 462 1112 1 310 451 578 847 1 539 194 720 959 1 1208 233 574 869 1 1359 94 472 619 1 1859 172 564 824 1 2245 11 414 430 1 2655 49 296 406 1 
+JPEGImages/PartB_00354.jpg 6 266 149 27 41 1 317 229 40 44 1 372 272 57 47 1 512 225 38 36 1 81 273 56 63 1 126 211 39 35 1 
+JPEGImages/PartB_00078.jpg 26 613 272 27 59 1 538 318 42 44 1 576 274 28 29 1 568 254 30 31 1 497 263 25 27 1 375 214 33 33 1 356 261 19 20 1 309 285 17 19 1 291 301 17 17 1 244 299 21 24 1 209 313 26 30 1 153 321 30 32 1 142 287 16 19 1 208 280 17 21 1 230 273 17 16 1 143 264 14 16 1 131 240 12 15 1 111 274 17 20 1 110 241 15 16 1 71 243 17 15 1 53 243 14 15 1 45 237 9 11 1 35 251 16 18 1 17 249 15 15 1 13 291 22 20 1 57 284 18 19 1 
+JPEGImages/PartB_02316.jpg 20 6 165 76 70 1 21 84 26 31 1 28 78 24 27 1 30 52 15 16 1 71 59 18 19 1 109 78 16 15 1 126 66 15 13 1 144 70 16 17 1 171 89 20 19 1 149 97 26 24 1 162 114 26 30 1 196 112 31 32 1 185 174 89 103 1 262 123 36 36 1 277 94 24 25 1 317 99 30 31 1 346 118 35 41 1 352 97 28 27 1 403 56 21 24 1 468 117 30 36 1 
+JPEGImages/PartB_01607.jpg 7 8 287 84 93 1 119 304 41 60 1 268 328 67 78 1 390 294 40 54 1 500 247 165 186 1 644 296 84 132 1 831 285 97 130 1 
+JPEGImages/PartB_01804.jpg 15 3 187 57 61 1 48 148 22 26 1 112 101 23 24 1 90 169 51 44 1 186 160 42 53 1 225 133 33 32 1 191 125 21 26 1 334 139 24 25 1 350 126 20 21 1 386 110 21 26 1 400 148 56 70 1 408 131 41 38 1 456 118 27 30 1 488 132 35 32 1 520 122 30 34 1 
+JPEGImages/PartB_01756.jpg 7 7 130 87 134 1 113 128 64 69 1 166 114 41 50 1 176 88 31 40 1 196 83 22 34 1 210 86 24 35 1 289 16 22 29 1 
+JPEGImages/PartB_00832.jpg 4 48 99 27 28 1 96 102 21 25 1 167 97 22 22 1 238 95 19 23 1 
+JPEGImages/PartB_00373.jpg 12 54 128 38 35 1 54 95 35 29 1 105 102 42 44 1 154 98 34 34 1 58 173 52 48 1 212 134 52 39 1 270 114 45 42 1 318 101 40 38 1 356 156 45 65 1 392 161 67 70 1 395 126 51 42 1 462 112 45 37 1 
+JPEGImages/PartB_01375.jpg 10 45 179 41 38 1 31 217 67 50 1 161 193 57 62 1 144 175 30 42 1 100 161 27 35 1 254 197 58 56 1 265 169 26 32 1 315 158 41 47 1 384 185 63 57 1 441 172 35 35 1 
+JPEGImages/PartB_00296.jpg 11 170 118 72 59 1 113 67 51 48 1 70 51 44 40 1 98 36 46 40 1 207 1 29 22 1 230 45 37 40 1 292 58 45 50 1 357 81 63 57 1 303 126 81 96 1 454 74 70 75 1 534 58 49 56 1 
+JPEGImages/PartB_01180.jpg 36 74 438 73 65 1 13 583 166 129 1 16 412 51 40 1 4 375 30 36 1 112 422 46 41 1 111 391 36 32 1 140 388 29 31 1 196 388 26 28 1 189 409 49 57 1 226 406 28 37 1 255 399 49 42 1 267 425 47 50 1 326 393 32 30 1 313 348 27 35 1 347 328 22 33 1 345 447 78 97 1 411 405 41 46 1 379 390 34 44 1 493 376 37 46 1 501 409 34 44 1 508 442 56 66 1 419 483 89 112 1 543 455 86 116 1 564 383 39 41 1 545 378 31 41 1 618 415 54 54 1 636 390 36 42 1 697 400 43 46 1 722 440 61 65 1 771 393 36 41 1 760 360 26 29 1 823 413 55 56 1 812 390 41 32 1 863 406 43 51 1 898 393 42 47 1 927 382 23 38 1 
+JPEGImages/PartB_01720.jpg 29 1 196 16 27 1 16 193 29 24 1 64 200 22 24 1 75 197 17 18 1 108 196 24 20 1 129 200 21 29 1 132 190 16 17 1 150 193 16 20 1 155 202 29 30 1 184 210 26 30 1 198 201 25 24 1 187 159 21 19 1 227 211 30 30 1 246 199 20 20 1 258 193 16 19 1 271 167 17 16 1 325 174 18 19 1 274 208 23 25 1 288 200 19 16 1 357 216 35 39 1 376 152 17 19 1 415 200 26 27 1 425 225 35 31 1 473 210 27 27 1 499 221 29 34 1 527 200 26 28 1 562 157 21 24 1 489 200 20 21 1 464 147 21 23 1 
+JPEGImages/PartB_02360.jpg 8 87 40 61 67 1 148 81 76 88 1 218 27 55 66 1 277 37 63 72 1 318 16 40 53 1 350 140 106 127 1 396 87 79 98 1 485 30 51 63 1 
+JPEGImages/PartB_01647.jpg 69 319 297 52 63 1 433 287 53 53 1 467 232 30 40 1 318 233 36 44 1 260 220 28 34 1 221 206 24 27 1 186 219 32 34 1 197 187 21 23 1 164 177 19 22 1 123 184 24 23 1 127 163 14 16 1 96 166 20 19 1 105 151 13 13 1 85 150 15 21 1 92 142 10 14 1 67 145 14 16 1 62 139 10 11 1 76 136 10 11 1 66 111 9 8 1 541 243 26 30 1 666 258 40 45 1 659 237 30 30 1 711 225 24 27 1 711 202 24 26 1 733 194 17 23 1 724 184 18 19 1 783 200 16 28 1 703 178 14 18 1 650 195 17 24 1 593 205 25 23 1 532 209 25 24 1 520 197 18 21 1 486 200 19 23 1 449 212 22 29 1 389 197 21 23 1 330 194 20 27 1 335 180 18 21 1 305 174 19 20 1 266 174 16 18 1 269 163 16 20 1 247 161 13 13 1 250 152 13 10 1 224 156 9 10 1 210 158 14 14 1 306 140 15 16 1 322 162 13 18 1 343 163 11 15 1 383 174 18 20 1 381 168 12 15 1 398 167 14 17 1 413 172 13 14 1 418 182 18 21 1 463 184 16 19 1 461 166 13 17 1 431 167 11 16 1 489 181 16 18 1 510 162 11 13 1 517 178 17 19 1 525 162 14 15 1 545 171 15 17 1 565 182 17 17 1 588 164 13 17 1 601 190 18 22 1 596 183 14 19 1 614 166 15 17 1 628 184 15 19 1 643 175 14 16 1 660 171 14 16 1 661 190 11 13 1 
+JPEGImages/PartB_01737.jpg 44 16 130 14 16 1 33 136 11 13 1 44 136 12 15 1 58 133 13 17 1 64 136 16 19 1 88 144 19 21 1 108 130 11 15 1 121 130 14 15 1 136 132 12 15 1 153 128 15 19 1 171 133 16 21 1 192 133 16 15 1 189 124 8 11 1 209 131 14 15 1 227 129 13 17 1 189 144 24 31 1 211 145 25 33 1 249 132 13 13 1 249 127 8 9 1 277 131 14 17 1 283 123 10 11 1 294 126 12 15 1 309 132 12 13 1 316 126 8 9 1 322 128 11 13 1 332 130 15 15 1 332 124 9 8 1 344 128 9 12 1 351 124 8 10 1 363 128 11 15 1 372 125 8 9 1 386 130 10 15 1 387 125 10 9 1 406 122 12 16 1 435 127 13 13 1 274 154 31 36 1 345 146 25 28 1 396 145 25 28 1 391 138 15 16 1 430 142 20 18 1 445 134 15 17 1 455 156 23 32 1 470 147 21 24 1 489 134 11 22 1 
+JPEGImages/PartB_00103.jpg 5 34 5 27 34 1 30 94 35 42 1 155 99 34 35 1 203 93 31 36 1 195 51 24 32 1 
+JPEGImages/PartB_00044.jpg 5 680 246 324 352 1 407 217 211 227 1 333 32 85 107 1 173 254 145 132 1 64 277 102 90 1 
+JPEGImages/PartB_00554.jpg 7 69 333 280 333 1 390 345 154 209 1 518 266 116 158 1 463 250 95 108 1 719 284 115 127 1 885 308 116 125 1 1166 279 271 365 1 
+JPEGImages/PartB_01973.jpg 6 41 55 60 61 1 145 46 38 39 1 186 28 40 46 1 242 46 31 37 1 278 18 32 36 1 345 24 50 51 1 
+JPEGImages/PartB_01353.jpg 14 79 2081 503 573 1 399 1784 427 447 1 1072 1681 370 426 1 1049 2001 837 770 1 1599 2041 540 513 1 1626 1731 456 440 1 1949 1627 340 467 1 2869 947 283 357 1 2896 1717 366 397 1 3066 1857 440 364 1 2996 2044 476 560 1 3918 1808 360 383 1 4079 1877 344 377 1 4379 1781 367 671 1 
+JPEGImages/PartB_01414.jpg 25 352 77 18 22 1 309 98 7 9 1 292 102 8 11 1 276 99 9 11 1 255 104 9 12 1 256 96 7 9 1 222 107 11 14 1 231 98 7 9 1 215 97 8 10 1 190 109 15 15 1 187 103 7 11 1 149 115 16 21 1 146 107 11 10 1 130 97 8 11 1 147 99 6 8 1 171 99 7 8 1 112 108 10 15 1 100 97 11 13 1 93 102 10 12 1 74 118 22 24 1 54 113 13 15 1 67 105 12 14 1 52 106 8 10 1 15 111 13 14 1 5 108 13 13 1 
+JPEGImages/PartB_00895.jpg 20 190 102 10 12 1 212 107 12 12 1 282 124 16 17 1 235 103 10 12 1 230 92 8 9 1 266 87 13 10 1 287 79 7 7 1 281 73 7 9 1 292 48 6 9 1 253 65 7 7 1 236 72 8 6 1 226 68 7 8 1 159 82 7 7 1 175 94 7 9 1 171 74 7 6 1 153 80 5 6 1 226 51 7 6 1 244 56 7 7 1 238 59 6 6 1 214 63 5 4 1 
+JPEGImages/PartB_02147.jpg 45 406 314 106 99 1 577 253 94 98 1 583 193 57 58 1 355 229 49 48 1 236 204 64 58 1 117 222 71 81 1 123 192 49 52 1 61 219 55 53 1 101 145 32 35 1 86 130 30 29 1 46 127 27 27 1 51 110 19 23 1 22 133 29 29 1 155 150 31 31 1 178 166 33 27 1 183 120 22 27 1 221 140 22 23 1 261 137 17 22 1 261 182 29 34 1 314 160 32 34 1 310 129 22 21 1 335 148 26 34 1 348 130 21 23 1 371 137 23 26 1 384 159 25 26 1 420 155 24 29 1 402 132 27 22 1 442 141 16 19 1 482 152 19 25 1 448 172 27 32 1 483 134 20 19 1 509 126 17 21 1 523 147 22 25 1 539 163 31 35 1 530 131 17 21 1 545 138 22 25 1 583 151 22 21 1 632 170 27 25 1 639 150 22 25 1 688 200 58 50 1 705 174 25 26 1 704 156 20 23 1 722 156 23 28 1 764 164 21 27 1 775 203 25 38 1 
+JPEGImages/PartB_01702.jpg 16 45 144 66 66 1 55 99 49 43 1 181 37 29 34 1 178 91 34 44 1 185 122 67 71 1 55 233 189 105 1 273 92 33 34 1 267 85 23 27 1 311 97 25 22 1 319 139 54 63 1 376 90 39 37 1 408 116 52 55 1 426 92 36 36 1 403 202 126 134 1 504 113 48 58 1 534 115 42 47 1 
+JPEGImages/PartB_01596.jpg 16 56 57 71 65 1 149 63 67 83 1 190 70 67 71 1 248 66 71 84 1 369 195 102 98 1 553 27 68 90 1 583 1 250 317 1 778 138 84 137 1 843 120 66 75 1 875 92 59 87 1 1027 11 194 266 1 1244 133 122 136 1 1259 56 85 77 1 1265 176 247 340 1 1424 80 84 95 1 1481 126 112 104 1 
+JPEGImages/PartB_01107.jpg 59 2 662 86 116 1 135 664 131 135 1 51 601 60 70 1 4 612 57 57 1 110 600 51 59 1 192 610 60 54 1 236 608 45 56 1 281 642 102 103 1 272 597 37 37 1 384 630 68 74 1 378 604 62 51 1 476 419 116 125 1 463 701 295 314 1 906 596 205 225 1 1078 610 95 142 1 859 356 154 163 1 862 515 33 43 1 850 490 41 42 1 794 500 38 38 1 794 533 37 37 1 795 582 64 82 1 689 578 77 110 1 683 533 34 40 1 1165 563 57 62 1 1089 571 50 51 1 1055 562 36 54 1 1103 504 43 47 1 1143 486 35 43 1 1184 516 37 51 1 1194 486 35 50 1 1235 463 34 47 1 1272 526 37 55 1 1218 508 40 62 1 1273 490 33 43 1 1315 473 39 54 1 1309 541 61 59 1 1309 586 64 91 1 1192 684 240 330 1 1222 632 100 119 1 1332 626 190 232 1 1374 541 103 129 1 1465 600 59 78 1 1515 544 92 126 1 1554 599 145 187 1 1655 562 102 149 1 1573 508 64 85 1 1424 496 33 49 1 1406 486 31 50 1 1454 482 42 50 1 1509 463 37 55 1 1750 714 250 339 1 1788 571 212 247 1 1746 548 48 74 1 1700 510 47 52 1 1752 496 47 62 1 1770 470 39 49 1 1851 471 58 63 1 1881 456 41 45 1 1962 463 38 51 1 
+JPEGImages/PartB_01356.jpg 48 79 353 105 111 1 146 325 66 70 1 239 352 92 93 1 328 309 59 66 1 397 334 78 82 1 409 303 49 54 1 636 324 55 78 1 681 306 35 39 1 666 293 27 30 1 702 285 31 36 1 718 282 19 25 1 602 278 22 28 1 580 293 31 34 1 538 291 29 32 1 512 277 21 29 1 492 289 30 32 1 468 283 25 31 1 440 294 29 27 1 434 276 21 27 1 405 270 18 23 1 390 275 21 27 1 369 272 18 21 1 355 294 33 35 1 347 282 25 26 1 336 275 25 23 1 316 264 16 18 1 312 282 19 23 1 276 269 16 24 1 289 272 13 19 1 245 293 42 45 1 275 303 30 42 1 263 275 12 21 1 240 282 18 20 1 242 269 12 16 1 196 266 13 21 1 182 290 29 33 1 174 266 16 26 1 154 268 15 19 1 151 292 25 37 1 118 294 46 50 1 111 267 20 27 1 47 290 45 49 1 62 272 26 28 1 87 278 22 30 1 100 274 16 19 1 48 272 19 19 1 2 273 26 33 1 13 265 20 17 1 
+JPEGImages/PartB_02240.jpg 13 37 192 36 38 1 96 137 15 21 1 87 183 25 32 1 107 200 45 53 1 146 184 24 31 1 178 185 28 38 1 245 217 40 50 1 250 184 34 36 1 258 127 18 23 1 361 167 17 25 1 373 184 25 34 1 408 169 24 24 1 2 264 82 100 1 
+JPEGImages/PartB_01974.jpg 4 89 1 39 33 1 127 20 50 61 1 165 22 101 119 1 246 74 102 126 1 
+JPEGImages/PartB_00787.jpg 6 1 66 35 55 1 121 47 50 69 1 117 18 60 58 1 229 5 128 145 1 384 2 93 183 1 208 26 82 119 1 
+JPEGImages/PartB_00149.jpg 12 3 284 38 51 1 74 270 42 43 1 126 212 23 27 1 169 220 22 20 1 124 283 44 44 1 214 212 21 25 1 253 212 21 25 1 296 210 23 23 1 344 206 21 25 1 471 282 53 49 1 490 259 30 36 1 548 271 32 41 1 
+JPEGImages/PartB_00454.jpg 13 46 318 31 39 1 129 282 26 27 1 114 240 26 26 1 173 233 15 16 1 186 222 13 16 1 235 240 26 21 1 258 227 20 19 1 218 328 40 34 1 118 447 48 53 1 298 441 59 71 1 262 285 23 25 1 295 251 21 23 1 335 240 21 23 1 
+JPEGImages/PartB_01884.jpg 5 8 157 48 55 1 123 132 39 51 1 219 138 37 39 1 300 137 25 28 1 358 190 63 76 1 
+JPEGImages/PartB_01475.jpg 5 58 175 71 80 1 153 95 40 53 1 208 86 42 49 1 243 157 66 69 1 333 59 53 62 1 
+JPEGImages/PartB_00061.jpg 5 44 148 88 107 1 143 45 116 147 1 234 510 143 169 1 419 360 172 225 1 482 130 68 83 1 
+JPEGImages/PartB_02263.jpg 12 1 77 39 65 1 45 78 33 40 1 81 103 26 28 1 124 89 36 40 1 154 111 21 29 1 184 108 27 30 1 203 124 23 22 1 217 117 15 18 1 260 115 23 25 1 297 108 23 26 1 347 118 17 23 1 444 96 52 55 1 
+JPEGImages/PartB_00177.jpg 16 312 726 79 95 1 462 671 104 123 1 580 710 93 107 1 716 692 114 111 1 932 683 89 118 1 994 635 79 109 1 1055 660 121 134 1 1282 746 121 128 1 1351 696 90 114 1 1473 777 134 141 1 1696 819 114 127 1 1976 678 140 159 1 2262 737 143 173 1 2366 646 105 150 1 2587 749 191 204 1 2771 669 150 164 1 
+JPEGImages/PartB_02028.jpg 20 21 68 12 12 1 91 59 10 12 1 96 66 13 14 1 105 62 50 57 1 158 89 63 79 1 154 71 22 28 1 168 64 24 25 1 148 56 14 15 1 200 55 32 36 1 225 58 13 16 1 240 62 15 18 1 259 62 15 20 1 274 60 16 16 1 289 59 18 21 1 270 106 50 58 1 351 151 60 62 1 310 42 34 41 1 358 68 18 22 1 340 69 19 16 1 396 43 38 42 1 
+JPEGImages/PartB_00003.jpg 10 361 140 36 45 1 512 328 76 74 1 361 372 93 74 1 252 300 59 61 1 163 291 49 49 1 86 346 97 98 1 122 277 42 51 1 95 258 37 42 1 1 299 46 63 1 84 238 22 27 1 
+JPEGImages/PartB_01931.jpg 12 16 110 22 26 1 47 124 42 47 1 61 110 17 22 1 74 108 15 22 1 83 113 25 32 1 93 102 25 28 1 144 121 46 56 1 185 105 28 38 1 171 99 22 32 1 231 149 55 62 1 311 100 37 44 1 329 114 45 47 1 
+JPEGImages/PartB_01636.jpg 18 158 272 34 40 1 175 248 17 18 1 111 243 21 23 1 97 242 12 12 1 158 236 10 12 1 167 221 10 11 1 190 239 8 10 1 202 240 8 10 1 227 248 12 12 1 237 244 10 13 1 257 240 11 12 1 272 237 11 13 1 297 237 11 13 1 310 237 11 11 1 333 248 13 11 1 347 239 14 13 1 364 247 11 12 1 390 242 12 10 1 
+JPEGImages/PartB_01954.jpg 9 9 253 66 61 1 197 232 36 37 1 186 222 34 35 1 343 218 24 27 1 443 213 30 37 1 476 215 36 48 1 504 235 33 44 1 524 233 45 59 1 586 260 79 82 1 
+JPEGImages/PartB_02232.jpg 9 64 304 86 90 1 2 327 83 115 1 153 232 69 65 1 372 102 42 57 1 423 167 29 33 1 508 160 31 34 1 537 168 34 37 1 602 162 31 35 1 635 169 36 37 1 
+JPEGImages/PartB_01411.jpg 15 13 286 60 65 1 30 267 48 54 1 84 312 62 66 1 192 261 47 46 1 231 312 119 88 1 303 285 70 68 1 357 286 73 73 1 398 268 38 43 1 423 292 63 55 1 466 284 40 53 1 431 251 45 47 1 408 161 51 53 1 499 263 49 60 1 537 262 44 56 1 577 263 23 54 1 
+JPEGImages/PartB_01451.jpg 47 135 372 219 249 1 371 542 315 296 1 876 289 141 161 1 1098 256 108 129 1 1179 233 91 105 1 1343 339 153 177 1 1405 298 117 134 1 1457 231 103 127 1 1378 170 53 65 1 1311 188 66 71 1 1220 201 82 87 1 1184 155 55 68 1 1104 157 65 79 1 996 201 73 87 1 876 211 89 93 1 878 132 53 64 1 956 145 55 65 1 1019 135 57 65 1 709 219 100 116 1 780 177 77 86 1 773 143 55 58 1 739 141 51 60 1 676 125 47 57 1 669 186 69 87 1 522 273 117 133 1 500 201 86 91 1 617 183 65 63 1 617 123 44 59 1 529 130 48 59 1 457 198 63 81 1 479 154 64 66 1 433 145 48 49 1 419 132 31 35 1 321 180 80 93 1 284 239 91 109 1 223 190 60 65 1 329 152 49 49 1 350 135 44 51 1 247 130 49 54 1 222 156 47 46 1 172 186 54 85 1 107 200 87 108 1 53 172 66 84 1 127 154 53 52 1 1 226 51 81 1 18 176 42 47 1 19 135 51 55 1 
+JPEGImages/PartB_00892.jpg 36 11 590 46 48 1 56 712 100 84 1 59 804 152 141 1 275 730 133 116 1 259 638 57 60 1 238 605 37 47 1 141 609 55 60 1 111 579 43 42 1 205 570 31 36 1 256 568 35 35 1 270 588 40 44 1 341 623 48 58 1 350 582 32 39 1 384 596 36 40 1 388 566 30 39 1 429 565 29 37 1 450 588 33 38 1 479 574 31 40 1 518 572 31 31 1 524 599 39 41 1 500 642 58 67 1 422 651 64 64 1 575 592 33 44 1 594 583 22 35 1 618 584 30 34 1 663 614 46 45 1 680 659 73 75 1 751 666 56 60 1 715 604 40 47 1 707 584 30 35 1 735 588 31 36 1 812 609 32 36 1 846 590 34 42 1 861 643 51 49 1 906 601 50 58 1 1135 462 76 77 1 
+JPEGImages/PartB_00088.jpg 20 4 354 43 38 1 119 475 90 94 1 188 425 84 68 1 314 398 63 55 1 365 380 39 38 1 408 375 36 35 1 405 353 27 30 1 677 356 35 40 1 711 377 50 50 1 739 411 53 52 1 724 458 85 64 1 869 477 107 87 1 909 568 131 136 1 927 382 40 47 1 888 369 36 41 1 860 359 30 33 1 1031 387 57 52 1 1022 347 40 40 1 1074 368 26 29 1 1139 373 43 42 1 
+JPEGImages/PartB_00850.jpg 13 65 55 55 68 1 129 24 44 49 1 77 1 40 31 1 169 110 87 107 1 232 54 65 69 1 199 1 44 28 1 279 29 35 46 1 311 2 34 35 1 387 23 35 55 1 352 118 70 98 1 417 63 56 62 1 424 5 34 34 1 538 23 47 51 1 
+JPEGImages/PartB_01928.jpg 35 433 135 22 22 1 320 198 15 18 1 294 187 13 13 1 285 182 13 16 1 241 198 16 19 1 213 204 26 27 1 218 194 19 19 1 241 185 10 13 1 256 185 10 12 1 215 183 8 9 1 197 190 15 14 1 181 195 17 24 1 172 187 16 16 1 169 177 12 12 1 163 225 29 28 1 131 196 23 25 1 145 190 12 16 1 149 182 11 12 1 147 178 10 11 1 136 177 11 11 1 125 185 11 13 1 115 174 12 16 1 108 170 10 12 1 93 168 9 12 1 98 186 16 15 1 69 207 23 28 1 65 181 24 24 1 85 178 14 16 1 84 172 10 11 1 45 177 20 19 1 46 168 12 16 1 27 170 12 13 1 20 181 16 18 1 13 171 11 13 1 1 197 25 28 1 
+JPEGImages/PartB_02336.jpg 10 2 155 17 28 1 47 156 24 27 1 64 183 26 34 1 55 272 37 39 1 2 361 42 70 1 127 382 105 116 1 129 182 22 21 1 141 192 24 26 1 187 203 26 26 1 448 206 37 46 1 
+JPEGImages/PartB_01418.jpg 16 1 186 35 43 1 46 204 55 62 1 48 173 29 27 1 62 159 22 23 1 123 83 16 17 1 91 190 32 38 1 118 172 25 29 1 128 159 27 26 1 253 155 21 22 1 277 169 24 31 1 308 176 32 38 1 332 170 22 27 1 313 155 20 24 1 362 205 48 53 1 395 178 32 37 1 411 150 26 27 1 
+JPEGImages/PartB_01314.jpg 23 317 55 24 23 1 332 70 34 40 1 380 121 42 35 1 406 151 41 44 1 251 164 61 73 1 251 228 66 62 1 250 121 30 25 1 231 98 24 23 1 251 79 21 23 1 266 68 18 21 1 201 72 19 17 1 155 116 31 33 1 179 98 38 30 1 125 167 41 38 1 34 218 66 53 1 40 155 47 37 1 65 125 34 29 1 94 95 28 27 1 102 85 20 14 1 121 69 23 22 1 87 69 20 20 1 58 77 26 26 1 2 97 26 28 1 
+JPEGImages/PartB_00577.jpg 7 264 656 478 497 1 664 156 228 211 1 1303 81 292 286 1 1839 165 350 463 1 2412 184 250 331 1 2992 45 275 353 1 3262 601 738 762 1 
+JPEGImages/PartB_01225.jpg 6 70 23 240 278 1 263 194 144 204 1 321 184 238 292 1 412 155 176 187 1 549 223 180 236 1 820 1 276 187 1 
+JPEGImages/PartB_02227.jpg 27 50 162 43 60 1 32 128 11 14 1 14 124 15 21 1 1 129 9 16 1 48 131 11 14 1 66 136 20 24 1 87 132 13 15 1 92 109 11 11 1 101 139 23 24 1 128 128 10 12 1 152 127 10 13 1 174 124 8 11 1 178 129 12 13 1 181 130 14 19 1 198 129 12 14 1 203 140 18 21 1 163 152 29 28 1 261 161 34 34 1 326 134 24 29 1 331 126 13 16 1 365 129 10 12 1 288 129 16 19 1 276 127 13 14 1 262 127 10 12 1 249 125 12 15 1 221 124 11 11 1 210 125 9 10 1 
+JPEGImages/PartB_00538.jpg 4 175 172 204 288 1 370 96 191 237 1 81 337 132 137 1 37 322 94 137 1 
+JPEGImages/PartB_01883.jpg 9 84 193 46 57 1 148 171 44 51 1 265 185 41 54 1 283 239 87 97 1 369 239 56 75 1 397 168 40 48 1 441 209 50 62 1 479 171 30 40 1 605 195 41 38 1 
+JPEGImages/PartB_00765.jpg 18 1 330 83 140 1 1 250 88 97 1 53 233 66 73 1 143 216 48 56 1 172 202 41 41 1 231 194 35 33 1 258 191 26 29 1 275 193 16 23 1 287 184 17 25 1 301 183 15 21 1 660 276 88 87 1 657 172 29 34 1 508 259 49 67 1 445 213 41 48 1 422 208 28 39 1 410 197 24 25 1 401 188 17 23 1 419 182 15 18 1 
+JPEGImages/PartB_00056.jpg 5 172 659 91 100 1 435 663 102 102 1 869 690 84 86 1 1476 735 172 174 1 1925 709 219 272 1 
+JPEGImages/PartB_00798.jpg 4 148 98 98 118 1 144 42 90 98 1 213 10 161 159 1 417 3 173 195 1 
+JPEGImages/PartB_00911.jpg 6 57 176 29 32 1 169 181 29 38 1 234 255 47 61 1 283 121 24 24 1 325 153 15 20 1 257 137 12 15 1 
+JPEGImages/PartB_00846.jpg 8 1 268 37 61 1 74 265 62 65 1 248 249 40 42 1 386 243 37 44 1 530 249 35 45 1 629 248 47 51 1 703 265 69 62 1 705 237 59 49 1 
+JPEGImages/PartB_02014.jpg 34 119 501 66 56 1 5 487 47 53 1 65 452 32 34 1 93 374 26 37 1 77 381 25 29 1 52 410 17 15 1 32 406 18 14 1 11 401 17 14 1 141 443 38 28 1 131 412 23 22 1 175 409 19 17 1 196 416 14 18 1 199 435 25 25 1 225 416 24 22 1 199 524 69 63 1 260 458 34 35 1 295 425 23 24 1 318 435 20 23 1 361 427 19 27 1 350 470 40 40 1 388 445 30 34 1 390 421 18 28 1 410 429 28 30 1 441 429 19 15 1 453 478 47 40 1 454 448 25 28 1 484 434 21 21 1 563 444 28 31 1 557 430 20 21 1 486 425 17 14 1 604 441 22 22 1 619 455 25 27 1 670 436 26 27 1 488 335 12 14 1 
+JPEGImages/PartB_01518.jpg 5 80 82 54 54 1 149 68 44 47 1 239 54 33 35 1 525 117 86 87 1 448 163 107 89 1 
+JPEGImages/PartB_01088.jpg 40 60 214 37 38 1 73 183 38 40 1 48 180 29 33 1 49 167 27 24 1 112 140 18 21 1 126 132 23 24 1 160 147 29 29 1 176 188 28 39 1 184 170 30 28 1 171 170 23 26 1 211 191 45 46 1 226 133 22 19 1 252 125 20 16 1 233 119 18 17 1 239 109 15 15 1 281 120 19 19 1 283 140 26 34 1 305 126 22 21 1 321 170 26 40 1 338 1 54 55 1 437 144 27 37 1 482 149 17 34 1 91 55 15 21 1 125 82 13 14 1 170 83 15 14 1 153 123 15 17 1 159 110 15 15 1 179 112 15 14 1 170 129 16 18 1 130 111 15 18 1 104 134 17 18 1 82 132 18 17 1 78 115 10 16 1 67 110 12 15 1 42 164 24 18 1 32 147 24 20 1 32 137 18 16 1 37 119 13 12 1 28 118 13 12 1 20 115 12 13 1 
+JPEGImages/PartB_00778.jpg 22 19 114 27 32 1 45 118 22 29 1 67 153 24 29 1 78 133 23 24 1 90 115 22 24 1 111 130 21 25 1 126 146 21 28 1 152 153 22 26 1 155 130 23 25 1 191 130 21 25 1 230 145 22 28 1 248 138 19 25 1 264 127 16 22 1 281 136 19 25 1 290 156 27 34 1 287 127 19 22 1 299 132 20 23 1 318 126 19 22 1 319 149 22 25 1 343 137 20 25 1 367 138 16 21 1 385 135 20 24 1 
+JPEGImages/PartB_00418.jpg 11 275 206 58 61 1 544 293 92 98 1 667 114 23 46 1 629 38 34 39 1 507 70 46 45 1 475 20 36 45 1 403 48 44 45 1 372 3 27 27 1 181 20 32 29 1 150 78 51 49 1 37 6 57 46 1 
+JPEGImages/PartB_00748.jpg 12 1 173 28 37 1 19 158 23 23 1 54 165 21 20 1 38 181 37 40 1 88 170 27 32 1 76 207 74 88 1 156 185 41 37 1 148 166 22 22 1 157 135 12 17 1 200 184 30 31 1 247 190 40 50 1 297 188 31 37 1 
+JPEGImages/PartB_02199.jpg 26 109 146 10 8 1 272 165 19 23 1 352 169 28 34 1 409 198 61 54 1 410 156 24 21 1 399 153 17 18 1 405 145 15 14 1 440 145 16 14 1 471 153 29 31 1 373 147 12 12 1 347 149 16 20 1 331 154 10 12 1 336 146 13 14 1 327 151 10 12 1 308 150 10 11 1 270 153 12 14 1 269 149 9 8 1 280 147 4 6 1 241 150 7 7 1 225 151 6 6 1 222 149 5 6 1 207 152 12 10 1 193 155 8 7 1 185 153 6 7 1 175 155 6 6 1 153 155 8 7 1 
+JPEGImages/PartB_00880.jpg 11 98 125 26 29 1 160 128 29 29 1 256 122 25 27 1 322 128 25 23 1 372 128 20 20 1 404 127 21 18 1 436 126 18 19 1 414 151 34 39 1 463 142 33 42 1 506 141 32 32 1 542 132 34 36 1 
+JPEGImages/PartB_01837.jpg 10 1 57 16 34 1 8 52 28 27 1 45 62 29 31 1 80 106 53 58 1 170 114 57 55 1 182 55 34 38 1 170 51 26 31 1 268 72 37 40 1 278 37 23 26 1 355 41 23 23 1 
+JPEGImages/PartB_01121.jpg 6 56 397 108 88 1 195 437 121 115 1 336 375 93 106 1 548 265 51 56 1 979 459 139 153 1 1042 408 81 94 1 
+JPEGImages/PartB_02352.jpg 7 128 139 21 21 1 183 126 27 24 1 201 115 27 27 1 242 128 29 36 1 269 110 37 36 1 331 115 48 54 1 436 103 61 66 1 
+JPEGImages/PartB_00298.jpg 26 1 132 46 74 1 36 122 40 48 1 56 109 25 29 1 72 104 21 24 1 138 69 14 18 1 121 90 41 53 1 103 121 67 81 1 195 96 38 47 1 277 121 52 67 1 290 104 36 42 1 282 90 32 36 1 247 88 22 24 1 233 107 22 29 1 166 102 16 19 1 162 88 16 16 1 172 80 15 18 1 184 93 17 21 1 329 91 17 25 1 356 81 26 34 1 384 84 20 29 1 372 101 40 47 1 428 123 58 71 1 410 83 8 11 1 428 77 9 12 1 467 81 42 52 1 545 90 45 85 1 
+JPEGImages/PartB_00294.jpg 22 1077 342 122 267 1 948 156 44 46 1 893 173 51 63 1 815 170 51 54 1 701 213 72 68 1 622 217 86 97 1 687 185 44 57 1 608 186 41 41 1 534 204 62 67 1 428 233 74 72 1 456 202 49 50 1 336 302 144 126 1 186 276 100 116 1 243 245 80 81 1 313 215 50 58 1 338 197 38 37 1 392 197 38 47 1 256 198 57 46 1 213 212 44 50 1 145 254 60 63 1 120 221 54 60 1 30 228 57 42 1 
+JPEGImages/PartB_01550.jpg 8 10 301 102 109 1 150 248 120 125 1 370 285 95 90 1 558 280 88 92 1 617 155 226 224 1 805 276 69 91 1 969 193 144 158 1 1123 116 77 264 1 
+JPEGImages/PartB_01993.jpg 11 164 127 126 147 1 49 165 100 115 1 362 222 70 75 1 388 211 47 54 1 422 254 42 49 1 477 207 70 67 1 652 247 37 46 1 706 246 52 39 1 752 251 36 38 1 767 230 31 32 1 629 265 32 37 1 
+JPEGImages/PartB_00920.jpg 13 3 407 34 84 1 44 431 52 69 1 89 403 72 74 1 165 412 68 61 1 158 219 67 61 1 289 382 63 71 1 382 387 53 71 1 497 381 52 65 1 565 385 53 63 1 673 399 58 67 1 761 372 73 74 1 977 218 153 178 1 1029 4 138 142 1 
+JPEGImages/PartB_02205.jpg 18 90 525 67 73 1 33 511 56 53 1 71 495 64 57 1 229 549 72 75 1 310 488 59 65 1 328 475 32 38 1 338 460 42 41 1 432 493 57 65 1 432 469 42 32 1 667 393 35 50 1 962 481 31 38 1 995 488 52 58 1 1049 446 33 32 1 1072 452 33 40 1 1125 488 53 60 1 1149 476 49 50 1 1142 529 82 93 1 1241 443 28 42 1 
+JPEGImages/PartB_01746.jpg 11 92 168 20 23 1 106 174 16 19 1 92 141 15 16 1 138 167 18 22 1 167 168 16 21 1 192 172 17 20 1 232 179 15 15 1 261 178 13 16 1 307 156 17 18 1 319 170 17 19 1 339 165 21 22 1 
+JPEGImages/PartB_01832.jpg 48 151 1029 141 180 1 344 1336 144 163 1 285 1346 93 144 1 195 1324 112 153 1 283 1468 185 192 1 385 1585 222 214 1 495 1714 324 337 1 346 2038 473 410 1 1419 1973 378 434 1 1078 1587 278 308 1 858 1536 222 237 1 727 1477 158 188 1 683 1390 122 144 1 605 1312 124 156 1 646 1241 105 107 1 795 1263 83 93 1 824 1307 86 105 1 890 1307 122 151 1 1073 1285 137 183 1 1314 1307 127 141 1 1485 1346 129 168 1 1456 1202 95 115 1 1595 1258 125 141 1 1614 1390 169 163 1 1751 1319 115 129 1 1795 1219 85 122 1 1992 1258 135 134 1 2005 1368 136 144 1 1944 1421 178 188 1 2388 1560 265 252 1 2336 1380 178 207 1 2358 1270 117 127 1 2356 1207 88 114 1 2200 1253 100 107 1 2168 1214 73 88 1 2112 1302 80 83 1 2488 1243 83 120 1 2456 1195 66 92 1 2546 1317 105 139 1 2588 1285 97 114 1 2678 1195 93 129 1 2744 1231 102 127 1 2844 1312 109 151 1 2883 1236 107 110 1 3034 1260 107 142 1 2946 1360 139 166 1 2873 1465 215 242 1 3071 1707 192 358 1 
+JPEGImages/PartB_01942.jpg 63 5 236 11 12 1 23 236 11 10 1 66 238 9 10 1 51 241 7 8 1 58 239 7 9 1 90 236 7 7 1 100 238 8 8 1 108 237 9 9 1 109 219 7 6 1 120 234 7 8 1 130 232 8 10 1 139 232 8 8 1 151 235 8 9 1 158 235 7 8 1 172 233 7 10 1 1 311 12 13 1 15 324 11 14 1 23 329 12 11 1 64 334 15 14 1 93 323 15 17 1 88 311 14 17 1 70 318 17 14 1 110 326 20 17 1 125 316 12 15 1 213 301 13 13 1 239 302 10 13 1 234 323 15 14 1 216 332 14 15 1 177 327 13 15 1 195 330 18 18 1 150 324 18 17 1 264 328 20 20 1 267 303 12 16 1 293 310 11 13 1 303 328 20 19 1 328 300 13 17 1 343 331 17 14 1 355 329 11 11 1 371 330 16 16 1 389 326 11 14 1 403 335 15 11 1 408 325 11 11 1 422 324 11 13 1 442 332 14 14 1 454 324 13 10 1 467 289 11 14 1 491 331 11 11 1 513 327 9 13 1 491 324 11 9 1 552 332 17 16 1 520 332 18 15 1 555 244 9 11 1 543 245 9 11 1 532 245 9 10 1 528 243 9 10 1 516 245 10 10 1 509 243 8 10 1 489 238 8 10 1 496 239 8 9 1 480 239 8 8 1 464 238 7 9 1 460 237 5 9 1 455 238 6 8 1 
+JPEGImages/PartB_00954.jpg 12 99 8 73 103 1 196 216 61 61 1 182 180 46 58 1 117 292 93 67 1 260 281 100 77 1 324 200 78 85 1 268 179 45 52 1 176 157 47 38 1 463 140 66 69 1 488 195 70 91 1 531 260 89 82 1 601 214 39 68 1 
+JPEGImages/PartB_01018.jpg 14 56 126 31 31 1 80 131 27 37 1 99 122 49 67 1 157 117 34 43 1 185 113 22 34 1 133 118 18 26 1 142 114 15 23 1 216 116 28 37 1 233 126 43 58 1 274 126 31 37 1 269 114 22 26 1 250 112 21 25 1 307 114 22 27 1 318 114 40 53 1 
+JPEGImages/PartB_00952.jpg 13 3 137 47 111 1 42 134 63 100 1 25 125 26 37 1 93 135 39 59 1 121 159 24 33 1 123 141 29 35 1 143 162 14 23 1 157 153 19 28 1 226 179 50 44 1 270 175 38 47 1 261 160 21 32 1 218 160 30 31 1 456 63 70 61 1 
+JPEGImages/PartB_00280.jpg 12 262 142 72 94 1 337 110 49 74 1 411 134 65 80 1 474 118 48 69 1 532 123 58 66 1 594 160 56 78 1 736 112 58 67 1 748 174 50 77 1 836 186 54 73 1 892 71 70 89 1 958 167 89 103 1 215 166 51 58 1 
+JPEGImages/PartB_02319.jpg 28 21 237 43 51 1 1 216 30 42 1 25 193 29 38 1 44 197 38 40 1 56 180 26 24 1 21 175 25 24 1 16 138 18 21 1 64 160 20 20 1 111 189 32 34 1 95 189 26 36 1 98 214 41 40 1 108 266 52 54 1 161 236 41 46 1 178 108 33 40 1 237 120 18 26 1 241 178 21 31 1 260 174 21 30 1 283 167 17 22 1 293 181 27 36 1 218 244 56 63 1 346 194 32 36 1 326 178 19 24 1 338 176 16 20 1 345 130 19 21 1 392 136 33 37 1 424 129 16 20 1 458 114 40 49 1 342 260 48 47 1 
+JPEGImages/PartB_00205.jpg 10 132 131 85 96 1 244 71 72 102 1 407 191 66 86 1 515 168 59 83 1 628 186 59 76 1 730 145 71 78 1 832 103 76 89 1 988 44 94 124 1 552 424 80 93 1 398 416 89 92 1 
+JPEGImages/PartB_02290.jpg 4 174 263 98 97 1 350 88 56 71 1 538 272 111 138 1 592 213 82 91 1 
+JPEGImages/PartB_02380.jpg 8 273 91 31 35 1 313 144 49 52 1 372 137 49 53 1 492 132 53 55 1 493 58 46 53 1 582 139 37 46 1 633 128 42 44 1 677 135 60 61 1 
+JPEGImages/PartB_01115.jpg 14 368 307 104 118 1 481 301 85 108 1 469 270 46 53 1 453 242 35 45 1 342 270 45 59 1 218 276 52 69 1 81 308 63 75 1 560 254 38 42 1 575 234 27 39 1 621 247 32 39 1 746 273 53 70 1 755 244 55 58 1 845 258 48 55 1 833 249 27 35 1 
+JPEGImages/PartB_00829.jpg 3 191 99 139 186 1 443 34 212 284 1 777 57 87 212 1 
+JPEGImages/PartB_00688.jpg 20 118 147 97 122 1 242 282 102 104 1 195 92 70 84 1 160 7 45 58 1 66 101 83 91 1 63 24 42 45 1 106 80 58 66 1 259 75 57 67 1 228 1 52 45 1 305 1 45 40 1 339 91 46 55 1 447 58 46 56 1 471 31 36 37 1 451 1 34 33 1 528 2 38 39 1 584 55 44 60 1 621 42 47 59 1 595 4 32 40 1 667 4 36 43 1 693 41 39 52 1 
+JPEGImages/PartB_01295.jpg 9 202 211 46 48 1 220 253 44 50 1 272 262 54 50 1 315 274 58 58 1 343 258 39 47 1 351 152 39 55 1 411 234 48 58 1 419 269 81 83 1 467 235 33 39 1 
+JPEGImages/PartB_00534.jpg 4 3 319 123 180 1 435 120 188 228 1 381 285 110 144 1 969 223 153 168 1 
+JPEGImages/PartB_02377.jpg 89 24 138 18 20 1 40 131 21 23 1 51 119 26 28 1 69 133 26 27 1 88 130 30 34 1 76 107 20 21 1 57 89 21 19 1 37 77 19 18 1 34 61 15 16 1 27 46 16 16 1 17 1 15 9 1 1 5 12 14 1 55 27 17 18 1 77 47 14 16 1 88 59 14 19 1 97 77 17 20 1 113 82 16 21 1 119 108 17 23 1 144 89 16 16 1 139 60 19 20 1 110 49 16 17 1 97 2 14 14 1 150 47 17 16 1 155 50 46 54 1 187 42 19 24 1 213 36 16 19 1 227 68 17 19 1 210 93 17 22 1 223 105 21 27 1 246 119 25 32 1 263 127 23 30 1 352 221 42 51 1 367 178 60 72 1 418 196 74 71 1 297 156 32 36 1 292 139 27 30 1 281 128 27 30 1 277 108 19 23 1 254 92 18 19 1 233 61 13 14 1 229 54 16 14 1 232 3 15 15 1 255 10 16 15 1 273 6 15 17 1 269 31 12 15 1 270 44 14 17 1 271 61 15 18 1 277 74 19 23 1 299 92 16 21 1 309 116 23 25 1 301 111 19 21 1 332 113 20 24 1 351 122 26 28 1 345 134 19 20 1 345 95 18 18 1 326 73 16 19 1 319 61 13 15 1 314 45 16 18 1 296 42 15 18 1 325 11 13 14 1 335 39 13 17 1 356 44 18 20 1 363 34 18 17 1 370 4 12 13 1 380 1 13 13 1 369 58 16 19 1 364 77 18 22 1 374 107 19 22 1 389 94 18 25 1 381 124 25 33 1 398 134 22 28 1 382 157 39 35 1 418 138 34 48 1 441 148 43 47 1 454 119 26 24 1 471 105 23 26 1 450 90 16 23 1 453 72 18 23 1 444 48 22 24 1 471 62 16 19 1 485 54 13 16 1 475 36 13 17 1 470 9 11 16 1 427 103 23 23 1 418 124 28 23 1 417 113 19 18 1 402 79 19 23 1 409 57 18 21 1 405 45 18 19 1 
+JPEGImages/PartB_01426.jpg 9 74 532 88 80 1 3 499 63 93 1 243 502 75 78 1 365 519 81 83 1 518 522 82 77 1 674 505 82 84 1 672 439 48 53 1 824 529 80 79 1 929 490 76 74 1 
+JPEGImages/PartB_01845.jpg 24 11 160 29 42 1 46 244 48 48 1 103 178 18 21 1 130 173 27 29 1 131 155 22 22 1 156 182 19 25 1 194 189 22 26 1 138 212 35 39 1 176 233 64 72 1 216 223 39 41 1 215 162 16 20 1 229 186 15 21 1 266 194 20 26 1 289 181 24 32 1 354 206 24 31 1 386 165 24 28 1 409 200 19 22 1 350 256 54 71 1 418 240 40 46 1 432 202 30 28 1 476 199 27 25 1 502 189 22 29 1 542 202 19 23 1 577 209 23 36 1 
+JPEGImages/PartB_00841.jpg 6 23 41 72 69 1 126 18 44 57 1 221 17 29 37 1 238 26 42 53 1 272 37 68 73 1 361 8 19 20 1 
+JPEGImages/PartB_02037.jpg 23 89 120 53 72 1 170 131 37 41 1 212 127 27 38 1 235 126 18 27 1 252 137 20 26 1 252 125 18 20 1 290 119 17 15 1 303 123 17 21 1 333 122 15 18 1 343 128 12 14 1 368 129 17 19 1 383 116 26 26 1 391 131 25 31 1 419 134 37 51 1 407 125 15 18 1 427 114 12 19 1 441 130 16 20 1 461 163 51 65 1 487 125 19 23 1 513 134 19 25 1 532 128 14 21 1 554 132 22 28 1 568 120 15 18 1 
+JPEGImages/PartB_00235.jpg 4 89 356 69 69 1 205 376 70 73 1 337 375 65 74 1 497 355 59 75 1 
+JPEGImages/PartB_00412.jpg 18 153 164 109 124 1 114 143 80 104 1 11 54 81 98 1 35 4 61 49 1 141 3 72 78 1 222 73 81 90 1 221 2 61 54 1 287 4 74 75 1 324 81 81 95 1 338 176 134 157 1 434 140 99 108 1 399 65 71 79 1 380 5 65 69 1 515 5 62 83 1 557 125 78 118 1 615 155 85 169 1 643 64 57 102 1 489 333 170 133 1 
+JPEGImages/PartB_02221.jpg 55 11 243 37 40 1 29 226 13 17 1 43 221 20 21 1 59 226 14 19 1 1 225 18 23 1 1 240 18 30 1 47 244 32 38 1 54 259 72 74 1 76 236 20 26 1 88 228 19 22 1 92 217 17 20 1 114 245 40 47 1 129 227 21 24 1 123 220 14 18 1 139 216 14 18 1 124 213 12 13 1 139 208 13 17 1 171 186 13 14 1 160 245 25 34 1 169 233 16 22 1 187 240 26 35 1 184 230 10 16 1 192 222 13 18 1 197 233 14 16 1 208 227 15 19 1 209 215 12 18 1 221 220 11 15 1 237 216 12 18 1 232 235 15 18 1 203 242 54 42 1 218 260 79 95 1 264 242 29 29 1 251 219 16 18 1 267 219 13 19 1 284 223 14 17 1 289 236 19 23 1 298 243 30 43 1 319 251 58 58 1 334 229 23 26 1 312 233 12 17 1 320 227 14 17 1 321 238 17 21 1 297 216 14 17 1 340 216 15 19 1 357 224 14 19 1 370 217 15 13 1 385 241 34 39 1 382 234 17 20 1 395 226 15 19 1 407 217 17 18 1 418 257 50 58 1 426 245 31 32 1 433 236 19 20 1 434 224 17 19 1 444 216 17 24 1 
+JPEGImages/PartB_02365.jpg 13 816 29 84 123 1 750 78 131 140 1 537 127 53 56 1 476 112 61 84 1 372 74 117 131 1 319 110 55 104 1 279 69 68 126 1 206 27 96 136 1 70 88 133 152 1 152 44 59 80 1 112 16 60 84 1 38 2 88 90 1 2 1 43 64 1 
+JPEGImages/PartB_01969.jpg 5 19 503 131 164 1 462 541 94 92 1 667 450 161 159 1 895 457 145 177 1 1139 491 109 136 1 
+JPEGImages/PartB_00581.jpg 4 1125 329 807 966 1 2280 512 845 993 1 3625 536 973 1159 1 4622 19 370 623 1 
+JPEGImages/PartB_02103.jpg 16 1 277 32 43 1 1 313 41 68 1 98 299 44 53 1 74 270 30 35 1 101 266 34 35 1 164 300 63 62 1 217 256 32 42 1 260 259 31 35 1 344 261 32 39 1 367 259 29 36 1 364 176 18 23 1 363 328 72 70 1 413 271 51 61 1 436 240 33 33 1 463 252 22 28 1 525 246 48 49 1 
+JPEGImages/PartB_01711.jpg 9 1 195 37 57 1 125 105 49 54 1 161 135 57 56 1 232 88 68 84 1 279 91 60 67 1 377 90 105 153 1 492 130 138 182 1 761 111 79 90 1 807 233 88 92 1 
+JPEGImages/PartB_01654.jpg 10 8 109 34 33 1 48 92 30 27 1 87 105 39 38 1 105 100 24 28 1 186 105 38 40 1 190 98 23 23 1 244 45 24 28 1 236 96 27 29 1 322 100 30 34 1 368 105 31 33 1 
+JPEGImages/PartB_01015.jpg 60 501 110 328 342 1 1004 599 138 167 1 1039 525 110 127 1 935 494 75 88 1 938 445 89 87 1 920 399 70 72 1 848 410 50 59 1 828 382 42 51 1 776 386 54 61 1 892 373 46 50 1 991 411 52 57 1 1034 423 71 87 1 1006 380 44 46 1 1038 364 37 50 1 1104 446 61 86 1 1210 459 98 121 1 1374 500 102 126 1 1365 586 111 179 1 1302 479 107 125 1 1333 438 76 73 1 1415 457 60 72 1 1381 401 47 53 1 1268 419 68 77 1 1288 384 37 41 1 1205 396 43 66 1 1229 371 36 42 1 1188 355 35 51 1 1173 405 41 41 1 1133 396 33 40 1 1146 380 31 30 1 1111 382 31 32 1 1101 355 30 29 1 174 660 164 181 1 6 664 96 216 1 130 474 110 136 1 201 553 157 145 1 293 428 121 113 1 66 425 91 93 1 1 447 71 123 1 146 365 49 68 1 213 400 62 65 1 281 389 48 61 1 320 376 51 52 1 370 407 67 97 1 443 393 48 59 1 379 359 48 61 1 407 349 43 48 1 483 344 39 41 1 344 348 48 47 1 286 341 57 52 1 245 359 43 42 1 225 322 41 33 1 201 347 35 36 1 151 334 35 37 1 131 315 43 36 1 104 350 44 40 1 79 332 35 49 1 12 340 47 51 1 14 305 39 41 1 81 396 53 49 1 
+JPEGImages/PartB_00116.jpg 14 134 397 91 84 1 300 417 54 56 1 304 350 50 48 1 440 543 113 122 1 536 440 62 79 1 546 393 46 57 1 543 358 38 43 1 628 374 69 64 1 653 324 48 58 1 685 430 65 79 1 738 527 116 129 1 780 266 37 34 1 1012 342 51 52 1 1118 337 50 55 1 
+JPEGImages/PartB_00917.jpg 36 5 358 58 71 1 3 325 48 51 1 54 251 32 33 1 23 210 27 31 1 80 209 23 26 1 127 207 22 23 1 112 231 32 35 1 182 242 22 24 1 184 206 21 22 1 226 229 21 27 1 235 201 18 24 1 274 228 26 28 1 320 228 23 22 1 363 261 23 21 1 305 258 23 23 1 275 266 28 33 1 231 277 34 37 1 167 280 35 41 1 136 395 72 79 1 267 412 75 96 1 380 374 60 59 1 265 348 55 59 1 215 332 45 44 1 352 293 36 32 1 388 278 28 34 1 410 258 21 24 1 420 247 25 25 1 437 268 24 33 1 473 272 27 27 1 504 268 25 25 1 530 253 22 30 1 660 228 22 32 1 729 324 39 41 1 839 292 28 31 1 900 371 64 73 1 932 337 44 55 1 
+JPEGImages/PartB_00950.jpg 39 38 127 47 48 1 97 105 37 36 1 63 56 23 25 1 87 66 17 21 1 41 43 13 15 1 5 45 18 24 1 4 22 13 16 1 124 54 14 16 1 153 63 15 16 1 173 60 16 18 1 206 60 11 14 1 194 81 16 18 1 222 81 16 16 1 238 74 12 11 1 259 76 12 17 1 281 79 11 13 1 142 154 48 52 1 276 195 51 70 1 240 136 17 21 1 299 134 22 27 1 332 138 27 32 1 353 178 35 39 1 411 228 73 76 1 500 208 41 52 1 556 151 36 35 1 433 131 22 25 1 475 73 25 27 1 446 89 16 18 1 404 89 12 17 1 395 76 13 12 1 373 77 10 11 1 366 91 11 11 1 352 87 14 15 1 335 79 11 11 1 325 72 13 18 1 311 78 10 11 1 308 91 16 20 1 299 102 12 18 1 283 98 16 18 1 
+JPEGImages/PartB_02329.jpg 8 135 50 56 65 1 219 50 46 62 1 269 84 46 65 1 310 87 37 56 1 352 76 58 65 1 411 63 43 56 1 445 97 49 52 1 505 85 49 66 1 
+JPEGImages/PartB_01976.jpg 3 25 219 20 28 1 107 197 16 19 1 327 189 25 28 1 
+JPEGImages/PartB_02099.jpg 7 86 285 41 37 1 195 264 27 33 1 122 331 80 69 1 256 297 40 44 1 327 275 24 27 1 392 213 18 21 1 562 276 30 30 1 
+JPEGImages/PartB_01046.jpg 3 227 198 183 211 1 497 176 188 243 1 625 147 313 377 1 
+JPEGImages/PartB_01998.jpg 10 50 1 115 110 1 8 60 35 32 1 28 52 22 26 1 215 75 31 36 1 227 68 24 27 1 246 76 26 28 1 290 92 20 23 1 281 77 20 20 1 311 94 20 24 1 358 65 92 119 1 
+JPEGImages/PartB_00342.jpg 25 45 1382 247 270 1 359 1335 208 231 1 192 818 239 270 1 53 771 195 206 1 3 860 114 219 1 564 782 223 278 1 731 1268 206 223 1 906 935 200 228 1 978 804 184 203 1 1148 804 155 170 1 1248 899 169 180 1 1187 1366 180 191 1 1434 1257 175 211 1 1431 788 156 214 1 1575 913 142 172 1 1737 1232 169 192 1 1948 1171 172 220 1 1717 782 128 178 1 1900 863 139 164 1 2067 804 131 184 1 2178 1141 172 194 1 2195 916 128 119 1 2387 1157 144 161 1 2484 854 119 145 1 2820 913 158 191 1 
+JPEGImages/PartB_01336.jpg 24 7 262 106 112 1 86 182 69 70 1 47 133 42 49 1 15 122 33 34 1 27 93 30 32 1 63 95 23 27 1 90 117 32 36 1 101 95 24 27 1 138 83 24 24 1 146 102 23 17 1 146 115 33 28 1 151 130 45 49 1 213 87 23 34 1 237 92 27 33 1 270 93 23 32 1 297 102 38 42 1 261 130 42 44 1 254 164 57 66 1 307 198 86 91 1 350 128 52 61 1 365 69 24 32 1 407 103 34 43 1 449 98 35 42 1 395 198 115 117 1 
+JPEGImages/PartB_01016.jpg 17 71 173 58 72 1 111 174 34 47 1 223 139 53 60 1 229 103 38 38 1 228 78 26 25 1 239 59 17 19 1 264 63 19 17 1 265 43 16 20 1 297 78 39 62 1 299 61 16 16 1 320 48 16 18 1 335 48 14 18 1 356 59 18 19 1 381 127 53 63 1 317 255 66 75 1 393 206 63 53 1 504 218 88 95 1 
+JPEGImages/PartB_02090.jpg 11 23 91 137 141 1 204 129 157 171 1 323 219 221 279 1 478 142 106 140 1 611 136 115 144 1 782 120 155 198 1 819 65 62 65 1 853 46 57 66 1 971 74 71 88 1 1089 67 75 77 1 1155 94 93 101 1 
+JPEGImages/PartB_01706.jpg 21 38 251 50 48 1 86 243 30 36 1 97 162 34 42 1 1 265 35 62 1 218 251 45 49 1 205 249 28 36 1 210 236 21 21 1 204 223 29 34 1 267 266 58 66 1 333 228 22 36 1 347 231 31 36 1 366 241 29 38 1 388 249 36 44 1 410 227 24 31 1 421 244 44 54 1 456 255 47 67 1 496 261 65 74 1 544 247 27 38 1 517 225 35 44 1 476 229 32 35 1 452 182 26 32 1 
+JPEGImages/PartB_02268.jpg 9 1 208 59 77 1 240 52 69 103 1 356 78 102 114 1 324 192 31 41 1 432 179 34 35 1 487 184 26 33 1 510 178 35 47 1 575 180 25 34 1 567 176 20 24 1 
+JPEGImages/PartB_02030.jpg 18 2 148 24 28 1 28 146 31 33 1 64 154 32 31 1 102 165 34 33 1 129 155 21 25 1 165 157 41 39 1 210 164 21 25 1 225 149 21 23 1 258 150 21 28 1 290 144 27 32 1 258 187 35 42 1 331 158 39 50 1 389 142 47 57 1 457 146 35 45 1 524 141 38 39 1 556 161 22 31 1 634 136 36 41 1 490 157 24 33 1 
+JPEGImages/PartB_00395.jpg 12 74 13 112 135 1 65 329 178 228 1 280 134 110 142 1 276 89 78 77 1 329 50 64 82 1 203 2 83 55 1 458 52 117 146 1 400 160 89 103 1 558 12 71 94 1 678 68 80 116 1 773 144 105 136 1 550 292 191 237 1 
+JPEGImages/PartB_01668.jpg 9 43 56 49 55 1 55 35 36 36 1 117 55 69 82 1 220 55 88 99 1 212 39 46 43 1 371 77 100 113 1 398 19 54 60 1 441 37 60 79 1 546 74 196 230 1 
+JPEGImages/PartB_01877.jpg 30 20 170 24 27 1 69 180 41 46 1 129 180 20 22 1 142 176 19 22 1 182 183 33 31 1 202 177 23 20 1 231 178 10 13 1 252 161 12 14 1 253 183 24 30 1 263 202 46 55 1 312 190 25 22 1 316 176 16 16 1 327 163 10 12 1 330 178 12 14 1 342 178 10 11 1 347 175 16 19 1 367 181 22 25 1 387 157 10 15 1 399 176 12 14 1 408 178 13 13 1 414 184 16 20 1 425 179 11 14 1 433 184 13 14 1 451 184 16 16 1 371 211 34 42 1 453 200 30 36 1 489 192 26 34 1 509 192 25 24 1 508 161 12 17 1 537 193 13 23 1 
+JPEGImages/PartB_02157.jpg 31 461 218 73 77 1 430 180 23 25 1 407 196 29 32 1 386 183 18 21 1 360 193 17 19 1 330 190 11 12 1 313 189 17 17 1 288 189 33 37 1 236 197 20 22 1 232 167 11 15 1 208 198 33 33 1 193 187 22 26 1 158 208 43 42 1 138 198 28 28 1 152 191 17 17 1 179 186 18 22 1 134 185 17 16 1 114 182 21 19 1 73 191 34 32 1 62 185 9 14 1 47 200 17 17 1 49 187 11 14 1 43 187 8 10 1 25 184 17 17 1 10 181 11 12 1 253 190 13 13 1 198 166 11 10 1 528 202 20 28 1 518 194 16 19 1 516 185 15 17 1 536 185 14 21 1 
+JPEGImages/PartB_00110.jpg 5 252 74 106 141 1 450 96 90 112 1 637 1 109 105 1 630 267 164 205 1 669 172 130 126 1 
+JPEGImages/PartB_01971.jpg 10 207 793 198 197 1 530 779 148 180 1 958 664 167 209 1 1345 761 169 214 1 1132 1030 287 347 1 1381 1222 447 551 1 1619 964 197 260 1 1823 753 140 209 1 1861 981 244 321 1 2170 730 160 187 1 
+JPEGImages/PartB_01482.jpg 24 32 104 36 37 1 25 78 21 25 1 14 77 18 21 1 7 73 13 16 1 22 71 13 11 1 98 63 14 18 1 112 74 18 19 1 111 65 13 15 1 128 69 13 15 1 144 66 12 12 1 147 79 21 27 1 161 69 15 19 1 170 64 12 12 1 176 68 12 18 1 185 76 18 20 1 194 66 13 16 1 204 85 23 28 1 208 71 14 16 1 216 61 13 13 1 233 76 18 21 1 235 65 13 15 1 249 66 12 14 1 269 75 24 28 1 272 64 13 14 1 
+JPEGImages/PartB_00054.jpg 50 1479 1324 34 32 1 2179 1395 73 70 1 2945 1326 64 66 1 3087 1524 97 90 1 2928 1675 161 129 1 2906 1558 117 122 1 2940 1473 73 85 1 2865 1507 78 88 1 2640 1595 134 131 1 2723 1524 78 93 1 2762 1495 68 70 1 2640 1504 78 100 1 2516 1533 71 74 1 2452 1487 88 98 1 2343 1463 68 73 1 2021 1690 188 161 1 1974 1577 105 132 1 2043 1558 124 107 1 2009 1502 70 73 1 1779 1470 88 71 1 1674 1551 103 100 1 1689 1507 66 61 1 1538 1541 83 76 1 1413 1629 130 136 1 1177 1692 178 183 1 1145 1595 132 107 1 1301 1480 90 100 1 1372 1463 73 88 1 1474 1473 64 68 1 1523 1463 51 56 1 1113 1475 78 73 1 1172 1451 61 61 1 974 1463 81 78 1 877 1519 85 71 1 626 1690 197 170 1 645 1582 122 108 1 750 1492 73 68 1 738 1451 61 61 1 640 1458 66 66 1 579 1485 78 73 1 411 1512 107 80 1 316 1602 127 115 1 404 1460 58 57 1 243 1519 83 73 1 150 1551 117 90 1 45 1602 129 115 1 77 1504 68 64 1 60 1470 53 56 1 228 1458 61 68 1 313 1453 54 66 1 
+JPEGImages/PartB_01987.jpg 17 29 141 42 47 1 117 116 47 55 1 160 103 31 37 1 227 87 24 29 1 249 97 26 41 1 267 73 20 28 1 290 99 21 27 1 282 126 48 62 1 310 84 20 29 1 329 46 19 27 1 361 84 24 33 1 341 94 33 44 1 398 119 60 88 1 428 96 28 33 1 445 126 30 38 1 410 37 21 28 1 518 24 27 29 1 
+JPEGImages/PartB_00610.jpg 23 20 99 60 71 1 71 81 53 56 1 139 126 46 52 1 161 239 62 73 1 193 100 53 62 1 160 63 41 50 1 241 71 42 45 1 264 126 47 54 1 299 93 44 54 1 340 53 35 57 1 376 78 38 46 1 405 97 52 54 1 457 85 29 50 1 491 99 43 53 1 524 69 35 45 1 530 107 55 64 1 583 80 38 50 1 605 104 47 55 1 632 62 43 51 1 657 94 43 58 1 723 103 44 56 1 462 230 58 76 1 335 231 50 68 1 
+JPEGImages/PartB_02386.jpg 8 45 106 30 39 1 96 111 34 33 1 183 104 35 43 1 267 110 30 37 1 305 122 31 36 1 364 119 35 35 1 433 115 29 37 1 512 132 28 36 1 
+JPEGImages/PartB_00755.jpg 4 49 33 50 52 1 58 16 31 32 1 178 45 27 34 1 251 1 90 63 1 
+JPEGImages/PartB_02209.jpg 14 62 138 39 37 1 129 124 21 25 1 94 127 19 18 1 114 109 15 20 1 151 113 22 23 1 145 103 13 12 1 163 68 9 10 1 190 103 14 16 1 223 109 16 14 1 195 114 15 16 1 264 107 13 17 1 53 112 9 11 1 77 106 13 14 1 12 110 12 13 1 
+JPEGImages/PartB_00317.jpg 53 402 166 21 22 1 144 252 27 28 1 93 282 30 29 1 88 259 27 26 1 22 299 38 35 1 45 282 28 28 1 21 255 30 32 1 122 327 52 54 1 206 392 98 88 1 348 501 163 148 1 415 340 81 110 1 467 320 46 58 1 496 270 35 32 1 507 247 26 28 1 459 249 21 25 1 435 285 33 32 1 408 298 34 35 1 433 270 23 28 1 393 263 25 26 1 350 286 31 35 1 288 314 38 46 1 321 328 30 37 1 236 310 51 60 1 235 287 36 35 1 304 282 30 29 1 318 255 34 31 1 555 254 24 23 1 566 266 36 23 1 574 285 30 26 1 578 313 40 46 1 602 324 57 57 1 665 357 94 82 1 824 444 176 209 1 923 343 77 107 1 791 315 55 56 1 740 337 53 57 1 832 310 33 38 1 789 279 38 41 1 768 270 32 34 1 706 314 41 47 1 665 321 44 43 1 643 296 29 28 1 635 278 29 25 1 632 250 27 29 1 618 251 19 19 1 720 237 25 26 1 730 259 27 24 1 709 271 30 28 1 720 296 33 30 1 865 236 29 32 1 926 255 31 37 1 932 244 25 25 1 979 237 21 31 1 
+JPEGImages/PartB_02337.jpg 28 16 168 26 22 1 26 195 28 27 1 58 218 33 29 1 50 246 44 38 1 78 168 22 21 1 113 179 23 27 1 166 201 28 27 1 192 231 35 34 1 129 335 70 78 1 313 292 57 61 1 209 155 18 15 1 246 115 18 19 1 243 168 17 19 1 246 153 16 17 1 287 159 18 16 1 297 179 25 21 1 323 148 17 17 1 355 144 16 14 1 375 150 17 19 1 407 155 15 16 1 355 173 24 21 1 381 199 29 33 1 442 189 29 30 1 438 161 23 21 1 468 155 21 21 1 524 228 38 39 1 543 172 26 26 1 569 159 23 21 1 
+JPEGImages/PartB_02121.jpg 26 13 248 44 53 1 60 259 51 55 1 84 295 59 67 1 102 234 29 40 1 95 202 25 30 1 79 190 18 24 1 125 188 20 25 1 183 186 15 21 1 192 188 21 28 1 211 198 22 29 1 225 203 24 29 1 213 239 37 44 1 313 250 46 46 1 320 223 33 39 1 288 159 15 17 1 340 188 18 22 1 349 201 29 31 1 370 194 22 31 1 397 208 24 30 1 407 187 21 21 1 356 273 56 58 1 422 230 43 44 1 488 244 44 52 1 480 219 31 34 1 485 197 20 24 1 445 193 23 24 1 
+JPEGImages/PartB_00543.jpg 29 17 100 28 31 1 1 94 15 34 1 64 106 36 39 1 85 131 71 54 1 68 88 15 17 1 83 84 14 20 1 108 89 23 22 1 127 85 15 20 1 148 97 22 23 1 157 80 16 20 1 178 85 16 21 1 189 80 11 13 1 202 99 23 28 1 221 87 16 19 1 214 82 11 13 1 245 77 13 21 1 251 75 12 16 1 255 99 32 36 1 282 95 19 22 1 290 90 14 15 1 288 83 12 10 1 291 77 11 14 1 327 75 13 15 1 341 79 12 15 1 344 88 11 19 1 354 91 22 30 1 375 79 12 16 1 394 87 16 17 1 451 32 36 37 1 
+JPEGImages/PartB_01565.jpg 6 31 461 170 184 1 94 394 143 162 1 177 389 132 153 1 687 59 84 132 1 772 340 136 131 1 1090 344 110 197 1 
+JPEGImages/PartB_00129.jpg 13 10 75 100 116 1 120 87 47 58 1 160 133 44 49 1 242 131 46 52 1 323 141 38 46 1 362 125 54 59 1 361 98 41 37 1 338 95 36 47 1 382 81 43 53 1 424 115 50 64 1 470 111 51 46 1 449 206 47 65 1 479 234 63 94 1 
+JPEGImages/PartB_01432.jpg 19 18 105 39 37 1 57 102 33 28 1 66 90 20 20 1 81 81 15 17 1 37 83 18 18 1 20 91 22 18 1 1 79 17 22 1 2 193 47 74 1 149 162 52 59 1 177 116 42 40 1 170 102 21 21 1 164 95 15 16 1 148 80 14 19 1 121 84 14 17 1 105 92 25 27 1 190 87 16 17 1 193 102 28 30 1 225 97 26 29 1 444 52 46 52 1 
+JPEGImages/PartB_02363.jpg 36 417 107 30 37 1 338 140 15 19 1 319 135 16 17 1 326 130 14 14 1 303 133 13 17 1 262 135 17 20 1 264 123 15 17 1 252 129 16 15 1 199 131 17 18 1 211 124 16 19 1 187 126 13 15 1 168 136 21 22 1 158 155 31 37 1 159 122 16 21 1 125 130 25 30 1 63 138 34 38 1 77 82 31 36 1 105 114 15 20 1 112 102 9 13 1 104 105 11 13 1 151 107 11 15 1 174 118 13 17 1 131 121 18 15 1 143 116 12 18 1 134 108 13 16 1 169 110 10 15 1 58 104 16 19 1 41 106 15 16 1 27 114 20 22 1 2 106 18 20 1 17 106 13 16 1 17 94 13 12 1 29 88 10 12 1 57 90 12 10 1 2 89 12 12 1 4 79 12 12 1 
+JPEGImages/PartB_01479.jpg 11 17 1729 190 222 1 231 1760 303 291 1 1019 1773 220 270 1 1036 1695 176 187 1 1385 1648 237 237 1 1785 1324 166 227 1 2117 1712 214 258 1 2339 1792 222 246 1 2592 1668 198 249 1 2810 1712 204 256 1 2983 1687 190 217 1 
+JPEGImages/PartB_01168.jpg 9 2 295 40 48 1 50 296 46 35 1 167 280 36 34 1 219 285 35 32 1 274 293 37 40 1 353 270 32 34 1 398 284 29 28 1 531 287 38 33 1 1115 219 38 45 1 
+JPEGImages/PartB_00277.jpg 11 76 93 32 29 1 120 87 27 32 1 150 83 26 24 1 175 79 32 29 1 212 72 40 36 1 257 61 28 36 1 285 67 23 35 1 306 70 19 28 1 364 69 34 46 1 396 84 23 27 1 350 59 26 33 1 
+JPEGImages/PartB_01604.jpg 9 50 534 95 113 1 197 523 39 44 1 229 551 51 48 1 277 544 92 87 1 326 447 30 35 1 413 541 38 41 1 499 541 51 52 1 505 523 41 38 1 606 540 49 55 1 
+JPEGImages/PartB_01571.jpg 4 312 1 298 308 1 597 144 126 151 1 710 151 89 129 1 894 104 140 183 1 
+JPEGImages/PartB_01416.jpg 9 8 84 74 76 1 92 68 35 50 1 151 124 99 88 1 180 52 56 63 1 257 68 47 54 1 360 61 44 48 1 402 49 34 41 1 447 77 57 69 1 491 51 45 60 1 
+JPEGImages/PartB_00301.jpg 11 403 333 46 50 1 315 355 41 49 1 236 361 48 52 1 237 341 46 43 1 147 339 45 47 1 180 335 27 35 1 198 364 19 27 1 217 366 24 32 1 91 341 58 67 1 33 321 44 73 1 1 225 37 172 1 
+JPEGImages/PartB_01239.jpg 8 65 347 136 181 1 143 208 126 134 1 355 168 66 81 1 422 211 94 114 1 481 272 136 182 1 578 177 60 82 1 691 245 94 100 1 865 267 157 182 1 
+JPEGImages/PartB_01745.jpg 4 36 1 78 67 1 265 1 42 37 1 279 13 84 114 1 420 1 48 45 1 
+JPEGImages/PartB_02056.jpg 26 136 200 11 12 1 122 265 25 26 1 149 252 19 17 1 121 251 14 13 1 100 257 18 15 1 80 253 13 13 1 92 287 27 30 1 48 299 34 31 1 64 273 21 21 1 50 257 16 13 1 44 253 15 14 1 9 263 15 15 1 4 284 31 27 1 314 249 15 18 1 326 264 16 17 1 336 268 28 23 1 358 250 14 15 1 376 247 17 19 1 377 279 27 30 1 406 249 18 19 1 418 246 16 18 1 453 243 15 14 1 449 283 26 25 1 471 267 22 21 1 488 254 12 16 1 488 245 12 12 1 
+JPEGImages/PartB_00510.jpg 6 16 67 16 15 1 56 54 14 15 1 117 81 16 21 1 184 92 25 26 1 247 126 20 25 1 263 110 15 21 1 
+JPEGImages/PartB_00329.jpg 5 20 12 42 99 1 93 44 71 88 1 207 33 56 74 1 284 40 50 72 1 347 41 44 95 1 
+JPEGImages/PartB_01393.jpg 29 1 167 39 50 1 2 212 56 88 1 80 181 47 42 1 96 152 28 33 1 49 145 24 23 1 16 149 32 31 1 70 134 18 18 1 96 126 20 20 1 105 136 17 14 1 124 129 13 17 1 125 153 20 22 1 145 147 21 29 1 160 148 18 22 1 146 136 17 18 1 168 137 14 18 1 183 145 16 18 1 214 103 23 32 1 279 132 15 17 1 287 140 14 16 1 287 156 28 40 1 299 146 13 13 1 306 135 15 17 1 328 146 18 20 1 334 159 23 24 1 360 163 34 39 1 348 142 14 18 1 366 145 15 18 1 382 140 17 24 1 356 201 42 49 1 
+JPEGImages/PartB_00985.jpg 5 27 147 151 129 1 266 200 81 94 1 622 83 78 84 1 665 193 78 90 1 878 66 171 131 1 
+JPEGImages/PartB_01563.jpg 11 132 171 94 83 1 312 139 46 55 1 328 177 61 70 1 423 195 90 85 1 579 32 50 61 1 517 238 135 162 1 724 194 96 103 1 984 154 57 71 1 1077 173 90 102 1 1010 257 134 148 1 659 534 268 245 1 
+JPEGImages/PartB_01502.jpg 10 34 114 19 25 1 57 123 22 22 1 103 102 17 19 1 134 110 17 19 1 158 124 28 32 1 167 96 15 19 1 193 96 18 21 1 245 137 33 43 1 239 107 27 30 1 290 62 43 48 1 
+JPEGImages/PartB_01689.jpg 12 129 174 38 50 1 149 246 60 75 1 131 247 40 48 1 264 247 41 63 1 293 243 26 37 1 322 246 33 33 1 365 249 34 48 1 389 243 37 59 1 416 234 63 81 1 493 243 44 42 1 546 243 24 38 1 560 252 46 52 1 
+JPEGImages/PartB_01934.jpg 12 513 81 33 37 1 268 140 19 21 1 220 142 24 26 1 203 139 15 22 1 165 151 24 22 1 163 138 17 17 1 115 141 22 23 1 59 142 17 13 1 43 198 89 84 1 1 161 23 46 1 27 147 21 19 1 3 139 14 21 1 
+JPEGImages/PartB_02085.jpg 9 275 387 117 137 1 495 478 58 75 1 546 480 56 74 1 634 517 53 61 1 743 495 54 64 1 794 361 96 119 1 902 431 88 106 1 1078 449 63 80 1 1138 414 125 142 1 
+JPEGImages/PartB_02154.jpg 98 1 381 30 45 1 85 367 20 24 1 80 343 25 26 1 116 361 21 27 1 140 352 20 25 1 147 370 26 31 1 171 379 44 40 1 184 353 20 25 1 212 347 21 23 1 224 339 21 32 1 246 341 18 26 1 267 353 15 19 1 277 336 19 24 1 280 361 26 31 1 259 371 29 30 1 252 403 46 46 1 333 371 31 30 1 314 345 19 24 1 309 346 11 16 1 340 340 16 22 1 359 362 22 28 1 384 374 34 35 1 408 358 22 31 1 433 371 31 38 1 419 397 49 55 1 387 339 18 25 1 377 335 15 20 1 401 328 17 19 1 442 354 26 25 1 474 364 25 37 1 514 372 30 40 1 511 350 23 26 1 519 329 23 28 1 505 324 15 24 1 499 306 16 18 1 475 317 19 21 1 474 347 21 29 1 462 339 17 21 1 447 325 16 21 1 543 348 22 28 1 561 335 16 24 1 545 321 15 18 1 571 317 14 21 1 582 353 29 26 1 586 369 32 38 1 625 359 26 35 1 606 329 24 33 1 653 337 20 25 1 682 360 32 38 1 685 345 22 23 1 695 321 15 20 1 675 328 13 14 1 671 336 15 17 1 736 350 24 30 1 750 346 22 23 1 744 334 13 18 1 737 324 14 15 1 779 322 14 20 1 784 331 19 25 1 798 348 18 21 1 814 326 16 22 1 832 332 19 27 1 826 321 16 23 1 864 315 20 22 1 872 329 22 24 1 907 328 18 25 1 940 318 15 21 1 744 400 44 58 1 818 373 41 52 1 891 369 29 43 1 916 366 37 53 1 944 353 32 33 1 892 422 52 57 1 965 438 64 56 1 995 389 44 49 1 976 374 30 41 1 990 359 29 31 1 1016 332 23 30 1 1032 324 25 29 1 1035 347 33 36 1 1073 329 20 31 1 1094 331 19 27 1 1102 315 16 20 1 1104 285 16 23 1 1162 278 19 25 1 1153 285 16 21 1 1146 332 25 29 1 1106 352 34 38 1 1080 379 28 49 1 1105 426 67 81 1 1168 379 48 53 1 1181 350 31 35 1 1189 331 27 33 1 1212 326 21 27 1 1205 303 21 25 1 1233 288 23 24 1 1272 276 19 27 1 1221 347 85 92 1 
+JPEGImages/PartB_00372.jpg 13 53 203 102 121 1 203 108 74 83 1 379 251 103 136 1 492 113 60 80 1 484 8 52 67 1 589 39 50 55 1 655 36 42 66 1 377 30 56 59 1 341 71 66 76 1 324 26 48 54 1 158 43 58 50 1 130 8 45 48 1 42 11 46 51 1 
+JPEGImages/PartB_01610.jpg 22 2281 819 227 239 1 1888 1136 112 134 1 1822 1158 92 146 1 1724 1180 156 173 1 1271 1290 217 241 1 1000 1304 180 237 1 773 1307 261 297 1 1073 1099 100 139 1 1149 1182 119 137 1 1268 1153 122 129 1 1410 1160 104 127 1 841 1099 117 144 1 729 1102 127 154 1 585 1156 156 207 1 666 1090 90 105 1 544 1117 114 117 1 427 1112 139 131 1 327 1226 168 215 1 95 1253 222 254 1 53 1121 174 171 1 110 1065 134 110 1 256 1068 112 105 1 
+JPEGImages/PartB_00062.jpg 18 23 131 35 30 1 1 129 22 30 1 52 127 40 33 1 121 183 60 52 1 160 147 39 37 1 208 132 38 33 1 257 127 28 28 1 256 159 59 48 1 319 133 27 30 1 325 123 22 20 1 388 120 24 22 1 419 136 50 50 1 430 112 33 37 1 491 116 40 39 1 546 147 44 35 1 526 133 26 32 1 589 186 51 72 1 589 107 31 22 1 
+JPEGImages/PartB_00213.jpg 24 91 180 32 38 1 124 158 29 33 1 155 174 26 30 1 192 162 27 34 1 243 174 29 36 1 299 196 32 35 1 357 163 28 40 1 426 90 27 33 1 453 75 26 35 1 481 77 33 37 1 399 81 27 32 1 392 61 26 29 1 365 72 30 38 1 340 65 25 27 1 312 75 27 33 1 287 62 29 33 1 263 80 32 33 1 235 64 26 31 1 207 79 27 31 1 179 86 24 24 1 150 96 26 27 1 133 85 19 25 1 110 90 25 31 1 57 91 29 33 1 
+JPEGImages/PartB_02370.jpg 13 105 349 119 130 1 57 336 106 114 1 197 354 79 109 1 269 461 58 68 1 386 467 61 67 1 413 379 99 103 1 463 367 108 120 1 554 157 92 114 1 627 339 160 174 1 760 341 75 102 1 819 372 80 105 1 877 329 149 162 1 1049 325 151 174 1 
+JPEGImages/PartB_00357.jpg 5 24 70 78 116 1 197 85 78 103 1 316 93 65 86 1 430 106 62 85 1 515 100 84 105 1 
+JPEGImages/PartB_01697.jpg 21 124 146 45 55 1 92 116 35 35 1 49 100 30 31 1 30 82 26 29 1 16 69 19 18 1 58 75 18 18 1 67 79 17 19 1 112 80 22 23 1 181 112 27 30 1 251 124 36 40 1 208 73 17 20 1 202 64 15 15 1 178 71 14 17 1 153 62 12 14 1 290 56 11 15 1 331 60 12 15 1 317 84 19 24 1 383 102 20 30 1 432 89 24 30 1 373 71 12 18 1 450 66 18 20 1 
+JPEGImages/PartB_01259.jpg 37 188 365 57 64 1 35 322 45 49 1 43 304 36 27 1 86 295 21 23 1 114 299 22 27 1 100 318 26 36 1 111 276 19 22 1 150 271 21 21 1 146 299 19 22 1 169 300 26 29 1 199 291 20 19 1 197 275 15 18 1 225 301 17 24 1 239 315 20 32 1 241 290 18 22 1 243 265 13 22 1 261 296 21 30 1 284 288 20 24 1 310 294 24 22 1 301 313 22 29 1 346 289 22 31 1 359 311 20 26 1 374 286 21 25 1 383 238 16 21 1 417 313 21 25 1 398 303 19 23 1 410 275 20 21 1 441 291 25 23 1 451 274 21 21 1 467 308 18 26 1 492 288 22 20 1 486 272 18 19 1 521 278 20 26 1 541 291 20 26 1 524 310 22 28 1 570 315 30 28 1 611 315 27 23 1 
+JPEGImages/PartB_02373.jpg 29 33 311 67 58 1 12 303 44 43 1 15 285 44 40 1 1 282 29 28 1 101 272 28 33 1 117 264 40 43 1 169 306 59 60 1 178 286 40 38 1 181 269 32 32 1 191 257 25 27 1 253 257 31 39 1 260 277 50 54 1 241 318 59 56 1 264 212 22 19 1 310 198 31 36 1 312 259 31 37 1 337 271 33 35 1 366 282 51 50 1 370 252 29 31 1 362 350 86 60 1 430 272 45 55 1 404 251 22 29 1 432 254 26 28 1 468 251 24 29 1 495 250 22 28 1 470 307 97 100 1 505 274 64 64 1 558 272 46 64 1 522 206 21 25 1 
+JPEGImages/PartB_00754.jpg 4 36 202 204 232 1 287 212 128 145 1 393 180 80 109 1 688 1 237 226 1 
+JPEGImages/PartB_00710.jpg 9 84 29 125 114 1 177 29 71 119 1 215 28 49 54 1 346 16 37 43 1 366 27 46 45 1 467 25 66 63 1 595 62 84 99 1 669 47 116 122 1 795 41 65 71 1 
+JPEGImages/PartB_00980.jpg 6 518 895 166 217 1 1001 921 164 171 1 1470 577 278 298 1 2384 556 239 275 1 2726 1158 156 193 1 2689 1034 176 165 1 
+JPEGImages/PartB_01383.jpg 13 240 388 71 73 1 300 412 86 102 1 454 400 57 82 1 507 411 85 93 1 632 372 69 72 1 682 403 75 83 1 646 454 114 118 1 817 369 83 69 1 942 266 62 54 1 890 430 92 101 1 1087 410 93 112 1 1190 528 172 161 1 1509 436 172 150 1 
+JPEGImages/PartB_01460.jpg 35 69 323 116 125 1 2 345 51 58 1 29 277 26 28 1 45 267 22 30 1 9 284 22 23 1 18 271 19 18 1 65 275 21 22 1 79 284 24 25 1 94 276 19 22 1 133 282 22 21 1 144 297 37 37 1 163 279 29 29 1 190 291 41 39 1 253 302 38 42 1 226 352 50 46 1 305 323 49 58 1 290 300 25 27 1 299 288 22 31 1 340 313 32 44 1 339 288 33 36 1 394 305 28 30 1 396 330 55 61 1 392 400 95 80 1 460 338 75 91 1 449 298 32 40 1 394 293 25 24 1 474 284 24 32 1 501 297 54 59 1 546 307 62 81 1 542 288 26 23 1 598 292 23 28 1 617 251 22 25 1 629 306 28 38 1 649 296 22 30 1 669 295 22 31 1 
+JPEGImages/PartB_00814.jpg 8 98 274 200 221 1 330 380 129 104 1 1 346 77 164 1 14 310 82 89 1 434 54 318 428 1 781 293 146 168 1 1021 268 179 186 1 1294 63 331 467 1 
+JPEGImages/PartB_00740.jpg 12 53 292 107 107 1 201 310 59 72 1 391 1 175 139 1 586 328 29 45 1 644 334 31 46 1 680 341 33 46 1 846 225 141 180 1 1000 360 35 53 1 1012 345 34 39 1 1074 326 55 77 1 1148 313 87 109 1 1232 322 67 89 1 
+JPEGImages/PartB_00381.jpg 111 1252 184 23 25 1 852 709 151 188 1 126 998 248 201 1 313 573 133 152 1 14 501 112 104 1 2 447 60 79 1 9 376 63 56 1 1 410 41 56 1 99 382 68 58 1 65 342 39 38 1 204 420 78 70 1 191 382 74 66 1 135 350 52 43 1 193 329 53 49 1 265 497 134 107 1 282 442 79 78 1 286 393 58 52 1 292 368 48 38 1 254 362 46 46 1 279 331 47 35 1 355 324 52 49 1 393 358 49 56 1 419 327 35 41 1 440 398 59 68 1 488 403 59 61 1 472 461 80 79 1 563 492 94 92 1 704 522 106 103 1 538 616 146 163 1 891 590 113 132 1 1151 635 104 129 1 1104 534 92 100 1 946 499 71 88 1 1302 584 102 106 1 1341 469 59 76 1 1452 534 72 77 1 1421 466 51 66 1 1517 513 56 67 1 1572 485 61 66 1 1636 466 50 58 1 1582 434 43 50 1 1669 445 50 50 1 1705 437 40 45 1 1772 390 25 45 1 1742 432 34 36 1 1644 345 35 38 1 1572 383 40 48 1 1544 347 35 38 1 1523 361 29 36 1 1489 334 30 28 1 1491 377 32 42 1 1462 387 31 38 1 1521 414 37 43 1 1486 443 53 50 1 1410 392 36 46 1 1378 405 41 36 1 1371 378 33 36 1 1404 329 27 29 1 1334 427 52 55 1 1314 401 36 50 1 1310 368 26 41 1 1302 351 36 32 1 1275 318 32 29 1 1231 394 46 44 1 1257 452 57 59 1 1176 461 64 68 1 1156 424 38 50 1 1135 409 35 33 1 1171 387 38 40 1 1212 335 27 38 1 1223 314 24 24 1 1249 321 27 31 1 1177 298 31 28 1 1144 338 28 30 1 1129 321 26 26 1 1092 367 28 37 1 1087 334 27 30 1 1072 299 28 30 1 1039 392 37 44 1 1045 377 34 31 1 1025 343 31 28 1 1014 309 24 31 1 975 304 25 32 1 934 337 29 29 1 968 352 24 30 1 909 357 41 31 1 924 414 53 49 1 918 443 53 63 1 777 429 73 74 1 1100 476 71 71 1 647 422 68 83 1 725 395 56 62 1 763 389 35 57 1 799 367 31 36 1 815 353 42 44 1 863 338 28 28 1 847 320 30 30 1 818 316 29 31 1 794 327 25 30 1 772 350 32 39 1 751 330 28 36 1 718 334 32 32 1 691 347 34 35 1 692 321 26 30 1 577 362 49 56 1 644 359 40 46 1 640 320 47 44 1 513 340 53 49 1 494 331 40 36 1 467 347 50 45 1 594 334 41 43 1 
+JPEGImages/PartB_01919.jpg 39 26 109 10 13 1 42 119 11 12 1 56 122 10 11 1 137 125 9 12 1 171 124 11 14 1 193 130 8 10 1 237 174 53 56 1 213 146 36 38 1 206 127 19 25 1 226 128 12 18 1 244 132 9 13 1 260 139 18 19 1 272 148 25 25 1 311 159 34 36 1 341 152 29 30 1 377 147 23 36 1 364 133 21 21 1 346 138 15 17 1 327 129 15 15 1 311 142 21 19 1 316 134 13 11 1 306 130 11 12 1 295 130 15 14 1 287 136 13 14 1 280 129 14 14 1 388 120 11 16 1 383 131 9 15 1 253 131 15 17 1 1 155 23 40 1 14 142 22 22 1 29 128 29 32 1 65 150 22 23 1 60 137 18 18 1 68 126 15 21 1 85 124 17 22 1 103 128 17 17 1 120 127 13 13 1 17 131 12 17 1 1 138 13 20 1 
+JPEGImages/PartB_00174.jpg 13 166 872 83 82 1 345 871 83 88 1 546 859 96 89 1 712 877 93 99 1 857 866 95 106 1 1029 943 116 108 1 1180 912 98 108 1 1298 911 77 80 1 1394 894 118 132 1 1549 903 106 119 1 1702 889 101 113 1 1828 903 89 86 1 1269 808 74 94 1 
+JPEGImages/PartB_01965.jpg 8 1 92 39 52 1 69 91 48 58 1 115 78 34 35 1 185 58 75 98 1 288 81 22 32 1 309 61 45 64 1 354 57 25 38 1 386 57 14 47 1 
+JPEGImages/PartB_00695.jpg 4 57 74 49 55 1 118 56 63 65 1 180 57 71 89 1 214 32 127 124 1 
+JPEGImages/PartB_02364.jpg 13 170 255 55 56 1 204 241 37 41 1 232 198 48 59 1 272 219 48 56 1 346 195 66 79 1 375 175 51 57 1 428 170 71 82 1 497 194 60 76 1 574 179 58 61 1 601 111 97 114 1 747 188 65 75 1 830 207 60 71 1 987 169 73 84 1 
+JPEGImages/PartB_02256.jpg 35 20 111 25 26 1 57 155 28 30 1 61 121 21 23 1 94 129 21 23 1 88 157 23 27 1 126 162 19 22 1 113 126 16 19 1 132 124 19 27 1 153 123 17 19 1 166 154 21 26 1 177 124 18 22 1 188 114 20 21 1 205 122 19 21 1 218 155 22 24 1 236 133 16 22 1 231 120 16 18 1 251 161 21 27 1 253 128 20 23 1 267 118 15 18 1 283 117 14 19 1 286 125 18 23 1 296 160 22 26 1 304 111 15 24 1 317 122 19 19 1 328 159 22 25 1 338 117 22 25 1 329 108 15 18 1 356 111 14 19 1 366 117 20 26 1 377 109 19 24 1 370 158 24 31 1 399 106 19 23 1 409 107 23 27 1 430 97 29 30 1 491 111 27 34 1 
+JPEGImages/PartB_00323.jpg 7 89 176 102 81 1 233 348 137 135 1 358 270 103 108 1 305 138 65 85 1 753 103 81 90 1 756 312 112 110 1 878 386 122 149 1 
+JPEGImages/PartB_00931.jpg 20 39 220 65 52 1 41 157 43 38 1 63 123 40 30 1 3 96 26 30 1 59 75 32 29 1 85 65 22 22 1 121 68 30 25 1 94 93 33 33 1 121 167 51 37 1 149 117 43 35 1 201 66 22 24 1 258 65 26 21 1 232 92 25 27 1 249 120 33 29 1 259 163 52 69 1 250 229 71 57 1 403 152 41 48 1 381 115 47 43 1 328 71 35 35 1 316 54 29 28 1 
+JPEGImages/PartB_01058.jpg 19 1 129 32 39 1 7 124 21 17 1 88 104 20 18 1 111 84 12 14 1 111 118 14 15 1 121 115 12 13 1 131 123 22 16 1 142 117 12 13 1 177 115 16 18 1 197 105 42 39 1 295 140 30 23 1 246 138 25 27 1 286 121 15 14 1 302 125 13 14 1 319 124 15 16 1 327 127 18 19 1 340 140 21 20 1 383 137 15 19 1 383 123 15 17 1 
+JPEGImages/PartB_00275.jpg 19 1 316 43 81 1 72 263 53 54 1 32 210 36 39 1 136 233 29 37 1 155 251 45 47 1 175 230 30 28 1 254 249 41 44 1 315 247 33 38 1 346 226 35 33 1 362 300 69 80 1 427 274 49 54 1 489 282 41 42 1 434 238 29 35 1 471 238 28 34 1 584 264 44 45 1 611 253 30 35 1 652 256 31 28 1 684 249 20 28 1 726 256 25 27 1 
+JPEGImages/PartB_02328.jpg 15 8 3 126 139 1 193 167 71 88 1 280 141 55 80 1 348 160 39 55 1 355 138 29 35 1 381 159 27 42 1 383 151 23 28 1 396 139 19 24 1 410 155 20 25 1 421 139 16 22 1 365 261 99 111 1 404 207 66 77 1 435 178 56 61 1 520 135 27 31 1 506 142 20 26 1 
+JPEGImages/PartB_01709.jpg 7 129 133 83 87 1 155 104 42 47 1 243 81 25 31 1 290 99 36 36 1 352 133 43 56 1 437 119 31 44 1 534 97 28 33 1 
+JPEGImages/PartB_01103.jpg 19 4704 1552 480 965 1 4476 1206 553 611 1 4008 1181 418 514 1 3301 1185 753 750 1 2890 2099 1164 1357 1 1554 1749 1268 1246 1 2390 1370 529 722 1 2697 1170 461 565 1 2062 1224 371 418 1 1697 1163 418 450 1 1179 1399 629 739 1 772 1352 440 529 1 322 1445 604 736 1 215 1288 414 443 1 233 735 236 250 1 601 745 236 257 1 690 1206 357 336 1 497 1106 307 343 1 258 1110 279 292 1 
+JPEGImages/PartB_01847.jpg 32 1 117 30 30 1 1 96 18 22 1 32 96 23 26 1 69 106 26 23 1 64 122 25 23 1 78 144 35 36 1 105 161 42 41 1 87 84 16 22 1 98 91 16 18 1 116 89 18 17 1 133 93 18 21 1 150 128 30 31 1 186 95 20 21 1 176 86 15 15 1 170 84 11 14 1 197 90 13 13 1 210 91 15 19 1 253 103 20 23 1 217 131 30 36 1 168 202 55 63 1 298 196 49 65 1 276 134 29 32 1 277 115 26 23 1 293 97 20 20 1 331 119 25 26 1 334 101 20 22 1 328 140 33 41 1 361 126 28 28 1 366 106 22 22 1 354 178 51 55 1 404 103 26 25 1 433 110 17 22 1 
+JPEGImages/PartB_01597.jpg 8 8 271 134 156 1 210 152 121 117 1 366 360 170 249 1 573 234 139 172 1 662 134 96 136 1 845 245 150 155 1 1028 416 189 250 1 1068 177 129 156 1 
+JPEGImages/PartB_01386.jpg 6 22 100 85 95 1 98 136 51 67 1 416 192 26 30 1 445 184 36 38 1 525 192 35 35 1 557 187 61 58 1 
+JPEGImages/PartB_00731.jpg 5 57 53 66 84 1 94 79 86 99 1 181 89 93 95 1 354 5 79 90 1 460 53 86 102 1 
+JPEGImages/PartB_01366.jpg 10 11 204 77 82 1 84 147 52 54 1 108 105 29 32 1 124 94 23 24 1 162 23 39 50 1 233 137 45 48 1 264 180 62 78 1 5 106 31 30 1 31 99 24 27 1 61 87 21 25 1 
+JPEGImages/PartB_02396.jpg 14 1 39 38 56 1 45 111 63 80 1 92 78 27 26 1 123 66 20 26 1 248 21 36 54 1 271 117 55 67 1 211 178 82 73 1 220 52 42 46 1 304 38 49 45 1 367 59 38 43 1 385 46 30 37 1 430 44 32 44 1 445 1 123 112 1 551 17 41 51 1 
+JPEGImages/PartB_01277.jpg 39 25 204 60 49 1 5 233 77 106 1 23 184 45 37 1 24 161 38 32 1 77 191 53 61 1 97 176 46 43 1 73 166 37 38 1 142 168 28 33 1 162 183 40 35 1 177 210 59 60 1 187 175 27 31 1 209 162 24 28 1 222 166 30 29 1 237 187 41 42 1 248 162 27 27 1 278 172 31 37 1 304 168 22 24 1 290 163 22 24 1 324 157 30 33 1 332 183 32 30 1 331 196 42 43 1 287 217 58 69 1 171 270 106 117 1 407 293 138 107 1 412 238 69 66 1 394 206 57 55 1 406 186 36 34 1 390 181 25 33 1 390 166 23 26 1 435 161 30 26 1 464 176 41 44 1 457 177 29 34 1 492 168 26 30 1 513 195 48 56 1 519 175 28 32 1 544 174 29 29 1 567 188 43 48 1 618 172 36 48 1 567 125 25 22 1 
+JPEGImages/PartB_02347.jpg 7 1 95 37 41 1 4 141 62 60 1 10 175 83 88 1 158 1 54 82 1 167 111 60 73 1 226 134 58 75 1 245 30 61 78 1 
+JPEGImages/PartB_02355.jpg 9 24 91 57 71 1 103 67 50 62 1 115 107 63 79 1 190 63 51 71 1 242 88 65 83 1 292 62 39 51 1 334 75 54 62 1 410 56 38 46 1 455 55 62 69 1 
+JPEGImages/PartB_00545.jpg 5 87 197 168 166 1 292 330 111 115 1 645 101 109 153 1 751 251 120 138 1 917 177 165 182 1 
+JPEGImages/PartB_01999.jpg 57 1 224 43 42 1 41 228 33 30 1 43 201 25 21 1 18 196 23 20 1 1 188 13 20 1 28 185 17 15 1 24 177 16 14 1 43 167 13 13 1 64 182 19 24 1 66 197 26 26 1 76 230 47 39 1 137 220 43 39 1 119 207 30 30 1 104 212 20 23 1 91 194 21 26 1 123 187 20 18 1 117 181 16 18 1 79 161 14 16 1 105 162 16 18 1 127 164 12 15 1 146 181 17 17 1 136 175 15 14 1 154 165 15 19 1 151 161 11 14 1 214 129 23 28 1 193 129 24 22 1 166 182 26 23 1 171 172 20 19 1 189 195 26 28 1 191 219 28 23 1 226 212 38 37 1 224 194 29 30 1 251 155 16 18 1 238 144 17 20 1 241 172 19 26 1 275 208 36 43 1 318 201 32 35 1 323 237 44 33 1 308 179 25 33 1 277 162 20 27 1 275 154 14 17 1 265 146 13 18 1 277 115 12 14 1 288 150 17 23 1 303 156 13 21 1 318 154 17 21 1 337 174 29 34 1 365 170 24 35 1 368 198 35 37 1 407 207 42 47 1 398 157 18 22 1 432 162 18 28 1 367 140 21 23 1 353 145 18 20 1 342 135 15 20 1 384 138 11 17 1 384 130 9 12 1 
+JPEGImages/PartB_01509.jpg 24 228 141 45 49 1 369 149 37 37 1 444 147 36 38 1 487 140 26 25 1 514 141 29 33 1 445 128 24 25 1 427 134 16 22 1 395 130 15 19 1 383 131 11 13 1 369 139 16 17 1 351 137 13 16 1 336 133 17 18 1 313 130 15 15 1 285 128 10 14 1 273 123 13 17 1 248 128 13 16 1 227 131 12 14 1 218 133 10 14 1 205 133 15 14 1 187 137 12 13 1 178 128 14 16 1 162 130 16 18 1 127 136 43 54 1 76 139 40 43 1 
+JPEGImages/PartB_01208.jpg 9 226 462 175 199 1 241 397 120 125 1 466 213 132 143 1 384 599 228 198 1 584 500 164 209 1 748 480 136 191 1 815 405 108 137 1 711 374 107 141 1 1083 512 117 190 1 
+JPEGImages/PartB_01337.jpg 38 116 224 48 61 1 179 194 28 27 1 114 166 20 25 1 96 175 17 14 1 83 170 13 13 1 69 167 11 12 1 43 175 21 25 1 29 170 13 13 1 250 212 43 53 1 262 182 17 17 1 253 179 14 16 1 252 175 10 10 1 247 171 10 10 1 218 170 10 11 1 225 168 13 11 1 208 179 16 12 1 184 173 12 13 1 191 168 10 12 1 177 153 10 12 1 155 170 9 11 1 142 165 10 10 1 289 178 12 15 1 334 190 19 26 1 352 184 14 18 1 361 178 11 15 1 370 174 11 14 1 391 172 12 15 1 394 183 21 27 1 428 182 12 16 1 440 188 24 28 1 451 179 10 12 1 350 172 9 10 1 335 172 6 10 1 470 209 36 40 1 486 192 24 26 1 505 186 18 24 1 530 181 18 19 1 532 197 22 39 1 
+JPEGImages/PartB_01500.jpg 41 23 286 56 51 1 20 258 36 41 1 105 239 30 37 1 131 264 38 48 1 149 243 38 44 1 206 264 49 52 1 168 320 61 67 1 204 234 24 35 1 259 235 26 35 1 277 250 32 40 1 295 312 73 65 1 331 270 50 50 1 360 256 29 40 1 381 239 27 34 1 351 228 23 28 1 310 228 27 33 1 363 214 25 28 1 415 232 25 31 1 447 232 20 30 1 456 241 25 34 1 467 252 32 39 1 455 284 46 60 1 556 262 43 48 1 582 235 34 36 1 603 227 23 27 1 523 229 27 33 1 648 234 31 37 1 664 248 34 44 1 718 242 29 33 1 785 224 26 38 1 456 428 108 120 1 619 329 80 96 1 759 284 56 77 1 788 274 48 54 1 751 438 166 150 1 865 353 89 89 1 870 242 47 56 1 922 227 34 47 1 955 288 28 64 1 953 256 31 46 1 961 223 23 33 1 
+JPEGImages/PartB_01030.jpg 42 41 205 118 150 1 6 273 57 76 1 158 294 45 63 1 193 301 48 63 1 247 289 44 59 1 310 291 48 68 1 290 302 37 38 1 346 297 29 40 1 394 318 38 43 1 385 307 29 35 1 432 314 33 42 1 464 312 33 41 1 492 300 25 34 1 513 294 32 47 1 545 304 35 43 1 570 281 34 39 1 586 308 32 40 1 607 316 33 43 1 574 342 47 55 1 634 310 30 42 1 656 324 25 35 1 667 292 42 47 1 647 294 26 31 1 669 342 49 55 1 712 309 40 50 1 758 336 50 58 1 738 294 28 37 1 764 294 28 41 1 788 280 36 54 1 817 308 28 36 1 847 305 39 51 1 837 279 34 36 1 890 328 48 65 1 890 291 43 50 1 923 282 29 56 1 983 244 62 81 1 1080 261 70 83 1 1030 289 32 47 1 978 296 32 45 1 466 415 55 64 1 959 553 86 98 1 1092 571 104 118 1 
+JPEGImages/PartB_02270.jpg 34 120 146 39 46 1 234 148 47 45 1 224 129 39 36 1 220 80 32 40 1 263 58 27 36 1 312 116 45 48 1 322 81 35 46 1 402 142 60 61 1 410 113 44 37 1 405 82 42 43 1 463 89 38 53 1 368 46 27 33 1 387 28 31 34 1 422 7 25 30 1 483 28 18 25 1 342 1 20 18 1 373 1 22 17 1 297 9 22 24 1 264 20 21 24 1 245 16 17 23 1 240 5 16 17 1 210 17 21 21 1 218 39 26 28 1 183 19 26 31 1 187 5 22 23 1 137 7 20 21 1 135 23 21 25 1 125 51 24 29 1 74 32 38 42 1 36 60 37 43 1 53 42 27 34 1 91 4 17 21 1 8 42 24 26 1 21 16 19 24 1 
+JPEGImages/PartB_00532.jpg 8 1 116 32 98 1 80 126 59 81 1 212 145 66 76 1 301 59 38 52 1 329 24 47 65 1 388 58 64 78 1 456 42 64 75 1 455 97 145 227 1 
+JPEGImages/PartB_01210.jpg 13 512 234 53 71 1 403 439 62 65 1 500 454 70 79 1 553 441 53 64 1 603 470 94 110 1 672 430 71 72 1 724 415 54 65 1 783 498 116 136 1 842 459 81 87 1 913 417 77 76 1 1 596 81 64 1 1 539 70 72 1 1 493 60 56 1 
+JPEGImages/PartB_01619.jpg 5 105 199 133 160 1 319 204 72 99 1 525 171 88 118 1 623 152 227 282 1 862 44 212 251 1 
+JPEGImages/PartB_02340.jpg 10 139 124 39 42 1 156 164 48 50 1 159 101 22 27 1 233 171 41 43 1 268 160 48 50 1 279 93 32 34 1 325 126 51 63 1 367 169 59 63 1 329 76 34 44 1 362 90 21 29 1 
+JPEGImages/PartB_00293.jpg 12 288 118 21 29 1 316 199 30 29 1 413 244 30 49 1 245 197 34 38 1 225 188 26 28 1 175 222 43 43 1 177 193 34 34 1 146 178 30 29 1 71 235 57 57 1 1 230 43 53 1 23 184 30 30 1 71 180 28 27 1 
+JPEGImages/PartB_00603.jpg 29 75 71 27 35 1 44 94 33 38 1 38 151 63 59 1 3 198 54 80 1 1 256 53 135 1 150 182 95 107 1 96 140 47 43 1 103 187 59 62 1 176 146 38 41 1 206 117 35 34 1 210 93 28 38 1 253 142 50 48 1 269 182 64 68 1 335 159 46 60 1 374 177 41 47 1 385 233 136 152 1 483 222 81 93 1 511 148 52 72 1 677 191 47 58 1 677 169 43 47 1 722 168 81 85 1 629 44 39 41 1 863 180 49 61 1 830 240 95 85 1 845 305 178 204 1 1005 195 76 94 1 981 177 50 73 1 1081 157 54 61 1 1165 171 68 73 1 
+JPEGImages/PartB_00052.jpg 5 994 1336 269 355 1 973 2212 542 582 1 2229 2301 579 645 1 3139 2157 466 583 1 3780 2467 604 641 1 
+JPEGImages/PartB_02385.jpg 14 29 104 25 31 1 49 93 25 28 1 73 92 26 27 1 85 106 25 30 1 114 107 28 30 1 163 91 22 28 1 198 106 24 26 1 230 91 23 29 1 250 115 21 26 1 278 96 24 28 1 307 120 24 29 1 331 116 24 28 1 314 85 24 31 1 373 87 30 33 1 
+JPEGImages/PartB_01830.jpg 37 2 154 32 38 1 5 123 21 23 1 54 123 20 21 1 70 113 16 22 1 96 135 26 24 1 63 175 47 44 1 111 117 21 23 1 117 104 17 19 1 148 115 17 16 1 159 108 13 16 1 139 142 26 31 1 153 132 23 24 1 204 140 28 25 1 189 121 18 19 1 191 109 16 14 1 212 107 12 14 1 219 113 15 18 1 240 119 19 20 1 233 106 11 14 1 237 103 15 15 1 267 107 12 16 1 276 111 12 13 1 276 119 18 18 1 284 140 25 28 1 224 162 36 37 1 297 128 17 17 1 300 110 15 17 1 325 115 13 16 1 333 120 14 19 1 340 113 14 18 1 354 137 21 26 1 373 126 21 23 1 365 114 19 21 1 390 115 16 16 1 419 120 15 19 1 439 119 14 19 1 445 46 26 31 1 
+JPEGImages/PartB_01723.jpg 16 87 71 27 25 1 128 105 21 18 1 119 101 14 17 1 160 103 25 24 1 185 103 14 12 1 211 97 13 14 1 203 105 20 23 1 242 104 33 33 1 225 100 14 17 1 296 103 12 15 1 301 114 15 18 1 351 113 18 17 1 347 103 11 15 1 373 116 39 42 1 398 100 17 20 1 429 116 31 31 1 
+JPEGImages/PartB_02255.jpg 19 346 222 60 54 1 188 356 65 71 1 248 376 60 68 1 256 475 115 125 1 134 436 107 108 1 46 392 84 90 1 8 434 74 88 1 336 376 72 80 1 392 438 106 97 1 489 406 78 93 1 567 499 126 139 1 644 459 101 100 1 622 369 65 72 1 724 381 69 67 1 736 337 65 72 1 912 371 67 80 1 901 350 62 68 1 941 523 167 149 1 1013 178 144 152 1 
+JPEGImages/PartB_00580.jpg 8 64 122 55 70 1 167 161 48 64 1 250 145 53 63 1 336 182 52 66 1 424 132 47 55 1 502 147 50 59 1 587 108 45 64 1 660 143 50 69 1 
+JPEGImages/PartB_01419.jpg 22 10 177 28 27 1 25 171 20 24 1 64 172 22 23 1 74 184 30 27 1 95 172 14 20 1 115 176 18 23 1 121 169 13 12 1 139 181 25 25 1 152 166 17 16 1 160 175 17 18 1 177 175 18 18 1 190 172 13 12 1 212 166 10 13 1 218 177 13 15 1 307 181 26 29 1 338 180 19 16 1 348 185 24 24 1 366 195 28 33 1 381 187 15 16 1 396 182 22 24 1 412 193 23 24 1 429 186 23 21 1 
+JPEGImages/PartB_01621.jpg 5 235 229 125 165 1 231 166 119 111 1 508 128 105 140 1 692 196 140 165 1 890 218 192 236 1 
+JPEGImages/PartB_00646.jpg 8 24 278 88 101 1 3 396 102 196 1 46 209 60 50 1 168 223 112 123 1 307 189 122 123 1 544 211 121 138 1 611 283 163 190 1 655 33 144 179 1 
+JPEGImages/PartB_00721.jpg 5 1 34 104 200 1 129 15 158 159 1 339 11 122 150 1 434 10 177 209 1 636 5 100 130 1 
+JPEGImages/PartB_02228.jpg 53 124 251 73 48 1 65 198 39 34 1 53 163 25 26 1 38 152 16 20 1 1 140 10 11 1 17 134 12 12 1 30 142 10 12 1 37 143 11 13 1 47 136 10 11 1 50 140 12 15 1 62 144 11 12 1 79 147 12 16 1 95 152 18 21 1 117 163 25 25 1 120 139 10 10 1 130 137 10 12 1 145 138 11 13 1 159 142 12 15 1 165 136 11 11 1 156 117 9 11 1 177 134 10 8 1 179 141 7 12 1 184 140 13 14 1 192 147 13 14 1 200 140 9 9 1 214 132 11 12 1 219 137 10 11 1 226 141 13 15 1 230 149 13 14 1 213 153 20 21 1 180 200 36 43 1 240 209 58 62 1 275 171 25 28 1 266 157 19 20 1 253 141 16 19 1 239 138 15 13 1 243 135 10 11 1 252 131 13 14 1 267 138 7 9 1 273 139 9 13 1 277 144 14 15 1 295 141 11 14 1 305 157 19 21 1 320 146 18 15 1 315 140 11 11 1 329 136 12 13 1 347 146 13 13 1 332 143 13 14 1 360 159 23 27 1 371 143 13 11 1 384 148 21 26 1 403 144 15 15 1 421 137 16 15 1 
+JPEGImages/PartB_00585.jpg 8 86 59 40 54 1 3 18 25 32 1 162 89 62 84 1 210 94 45 74 1 252 1 19 16 1 317 1 32 23 1 327 64 36 43 1 317 119 83 118 1 
+JPEGImages/PartB_00413.jpg 5 617 179 86 95 1 725 259 169 187 1 986 277 132 154 1 1133 274 86 103 1 6 186 177 273 1 
+JPEGImages/PartB_00904.jpg 4 1 130 27 33 1 192 137 35 38 1 307 121 36 38 1 419 130 31 38 1 
+JPEGImages/PartB_00237.jpg 7 37 102 27 30 1 88 64 44 42 1 197 96 28 29 1 267 99 28 33 1 327 53 45 37 1 404 97 33 34 1 137 106 23 29 1 
+JPEGImages/PartB_00888.jpg 8 29 281 105 149 1 126 246 86 95 1 325 243 102 104 1 465 317 159 222 1 524 87 85 96 1 747 226 91 110 1 895 316 101 129 1 1144 284 132 192 1 
+JPEGImages/PartB_01072.jpg 45 33 161 35 36 1 62 133 23 22 1 1 101 8 16 1 94 137 24 36 1 94 117 19 21 1 91 101 20 18 1 118 105 16 17 1 123 142 34 43 1 155 137 24 30 1 140 125 17 19 1 164 120 20 19 1 162 100 18 20 1 192 116 18 23 1 204 120 19 24 1 219 121 20 22 1 213 100 15 18 1 232 76 17 18 1 207 149 26 34 1 226 143 21 27 1 259 142 25 29 1 278 148 30 36 1 288 123 18 23 1 269 114 18 23 1 252 114 19 24 1 263 81 17 21 1 285 81 14 18 1 302 85 13 19 1 311 117 18 22 1 304 105 18 21 1 334 86 14 16 1 345 113 18 21 1 325 129 17 26 1 340 141 21 30 1 366 151 31 36 1 357 91 17 18 1 348 76 13 16 1 375 67 17 20 1 387 88 17 21 1 370 115 16 25 1 385 122 17 24 1 409 135 23 31 1 434 159 30 36 1 459 141 22 25 1 443 116 23 28 1 499 158 36 39 1 
+JPEGImages/PartB_00030.jpg 32 926 526 68 69 1 807 302 32 32 1 639 317 28 31 1 541 338 29 31 1 528 496 89 93 1 645 452 70 74 1 758 417 44 47 1 819 434 38 31 1 841 424 31 25 1 862 401 35 30 1 720 396 28 34 1 672 394 34 41 1 603 403 45 41 1 521 419 49 44 1 420 448 51 52 1 309 567 100 98 1 35 483 83 75 1 283 488 56 58 1 220 441 52 47 1 323 426 44 45 1 397 406 36 39 1 489 393 33 37 1 543 387 31 30 1 454 384 32 31 1 384 388 27 28 1 297 396 46 35 1 268 322 32 39 1 203 383 27 28 1 104 423 48 46 1 87 387 41 41 1 22 398 28 33 1 1 428 19 34 1 
+JPEGImages/PartB_00053.jpg 162 128 935 177 173 1 435 985 147 141 1 278 867 136 132 1 146 778 116 118 1 7 849 89 118 1 107 733 100 91 1 44 687 88 105 1 1 617 63 95 1 23 508 34 32 1 32 526 41 48 1 87 535 36 41 1 116 549 53 45 1 157 571 64 57 1 262 624 68 59 1 378 628 70 71 1 485 660 68 68 1 553 687 77 71 1 637 694 50 64 1 703 742 77 81 1 805 749 75 72 1 1021 835 98 107 1 1162 810 114 114 1 1260 831 100 86 1 1464 862 77 107 1 1553 899 86 88 1 1644 874 91 107 1 1516 828 66 64 1 1603 819 50 75 1 1691 840 55 59 1 1398 790 78 63 1 1266 749 69 70 1 1187 724 64 70 1 1119 717 70 50 1 1085 781 56 72 1 953 796 79 73 1 964 726 57 55 1 905 724 55 50 1 2126 933 102 111 1 2307 899 89 93 1 2269 903 54 78 1 2448 915 75 79 1 2528 906 52 70 1 2564 867 41 66 1 2605 926 68 73 1 2757 951 59 64 1 2698 915 53 63 1 2921 921 45 57 1 2980 915 41 54 1 3016 901 37 43 1 3046 935 45 46 1 3076 894 43 37 1 3148 931 52 54 1 196 533 34 34 1 112 515 52 38 1 230 567 57 52 1 323 581 55 50 1 435 592 36 50 1 394 571 43 55 1 453 565 36 43 1 537 567 39 48 1 621 606 50 47 1 580 599 43 38 1 776 665 56 54 1 735 624 36 38 1 889 658 43 38 1 916 640 46 50 1 941 633 37 36 1 869 615 45 41 1 821 646 48 46 1 764 578 48 37 1 1808 901 77 89 1 1959 914 76 86 1 2087 921 53 75 1 2227 851 66 69 1 2187 893 55 69 1 2158 849 62 58 1 2069 850 57 65 1 2027 849 46 61 1 1960 848 66 61 1 1887 846 53 67 1 1840 853 64 73 1 1819 815 57 51 1 1885 809 52 47 1 1968 804 41 43 1 2023 793 49 55 1 2099 831 42 55 1 2253 812 43 53 1 2315 822 51 60 1 2342 787 36 39 1 2438 843 47 55 1 2392 836 35 49 1 2486 829 36 45 1 2527 797 31 38 1 2562 816 35 45 1 2790 922 41 52 1 2809 896 42 45 1 2844 887 47 52 1 2756 856 44 47 1 2647 862 49 57 1 2093 783 43 38 1 2287 777 38 32 1 2256 779 33 32 1 2626 839 37 39 1 2687 859 29 41 1 2405 884 52 59 1 1430 720 57 64 1 1474 785 46 52 1 1539 799 65 69 1 1584 764 61 58 1 1562 735 59 57 1 1710 797 54 51 1 1754 761 35 49 1 1799 776 39 39 1 1841 770 36 44 1 1888 771 43 40 1 1892 740 36 33 1 1937 726 29 38 1 1907 710 23 31 1 1883 718 20 34 1 2037 730 28 22 1 2087 752 27 34 1 1830 741 32 38 1 1820 707 29 37 1 1774 743 26 30 1 1677 775 52 54 1 1653 724 47 40 1 1680 694 29 33 1 1646 687 25 26 1 1536 739 37 48 1 1511 707 32 38 1 1547 687 22 34 1 1575 680 25 28 1 1604 695 24 31 1 1481 619 28 31 1 1442 693 33 32 1 1474 708 32 32 1 1368 681 30 36 1 1397 691 40 42 1 1376 739 39 50 1 1327 740 44 55 1 1333 706 36 38 1 1301 680 29 34 1 1321 665 29 29 1 1347 655 19 33 1 1243 702 43 55 1 1217 676 38 36 1 1267 661 38 38 1 1190 632 32 25 1 1169 690 39 51 1 1159 657 38 43 1 1100 682 45 33 1 1138 647 26 37 1 1138 618 33 37 1 1722 725 27 33 1 1770 703 24 30 1 722 694 62 64 1 675 674 51 63 1 696 638 41 44 1 714 596 30 35 1 658 579 42 43 1 2148 773 32 40 1 2189 769 32 39 1 
+JPEGImages/PartB_00355.jpg 23 28 182 58 67 1 1 234 47 49 1 64 240 37 44 1 92 242 44 47 1 123 237 26 42 1 140 255 21 30 1 150 256 21 38 1 199 261 27 32 1 220 257 32 40 1 249 263 27 28 1 290 262 21 28 1 303 258 25 36 1 340 257 30 35 1 367 256 27 29 1 382 227 40 45 1 475 243 52 54 1 637 254 23 36 1 660 266 23 31 1 689 222 31 38 1 727 245 43 47 1 783 204 42 58 1 804 189 62 70 1 893 187 58 85 1 
+JPEGImages/PartB_00004.jpg 8 159 209 177 251 1 305 173 132 213 1 405 152 142 168 1 546 167 117 143 1 878 65 139 138 1 881 213 109 146 1 944 147 162 208 1 1075 110 125 277 1 
+JPEGImages/PartB_01548.jpg 10 5 116 26 37 1 13 97 26 29 1 39 105 27 40 1 39 85 19 20 1 63 90 22 19 1 70 105 22 29 1 87 114 27 30 1 92 90 19 24 1 154 94 23 19 1 277 70 37 43 1 
+JPEGImages/PartB_01690.jpg 68 111 191 82 83 1 38 176 51 59 1 16 148 38 43 1 6 135 29 31 1 4 113 18 20 1 21 115 18 21 1 34 119 23 22 1 45 113 20 24 1 58 131 22 25 1 68 117 19 19 1 82 112 15 16 1 103 123 18 20 1 104 115 15 18 1 94 138 28 31 1 122 140 31 34 1 141 128 25 25 1 139 121 18 18 1 172 128 22 28 1 166 112 15 19 1 177 111 12 14 1 188 122 16 20 1 195 116 14 16 1 203 110 12 14 1 219 117 14 18 1 237 120 16 18 1 205 135 32 36 1 186 159 42 46 1 286 170 70 75 1 304 147 35 39 1 290 134 24 35 1 261 118 18 18 1 234 112 11 11 1 244 109 13 13 1 271 109 10 11 1 300 116 16 20 1 294 125 14 15 1 307 112 12 17 1 316 112 8 11 1 369 141 31 45 1 411 156 56 67 1 362 251 137 117 1 499 124 21 28 1 541 118 12 16 1 527 108 15 13 1 518 112 12 12 1 468 125 25 26 1 480 115 14 17 1 489 109 11 15 1 471 109 13 17 1 460 112 12 16 1 446 121 17 23 1 425 131 26 26 1 405 121 22 23 1 423 113 15 19 1 435 113 13 14 1 426 109 11 10 1 433 104 11 9 1 409 107 12 14 1 400 112 11 12 1 388 117 14 18 1 375 115 12 15 1 357 122 18 23 1 335 125 15 20 1 341 110 14 20 1 329 102 12 15 1 363 108 12 13 1 377 108 11 10 1 388 106 11 14 1 
+JPEGImages/PartB_02247.jpg 30 1 513 112 131 1 3 444 77 62 1 12 417 55 52 1 76 479 89 107 1 146 404 62 82 1 116 391 57 68 1 183 429 74 85 1 218 480 101 105 1 313 393 60 58 1 370 401 62 77 1 419 477 82 81 1 446 443 73 72 1 454 388 52 56 1 502 388 44 55 1 394 558 120 130 1 584 256 108 121 1 698 363 51 58 1 732 367 32 48 1 764 391 49 67 1 803 377 46 48 1 823 415 65 92 1 854 401 72 78 1 936 364 48 49 1 997 329 49 50 1 1036 401 73 88 1 1052 381 60 54 1 1075 355 42 50 1 1108 406 69 74 1 1041 514 153 182 1 1079 468 117 104 1 
+JPEGImages/PartB_00504.jpg 16 60 100 25 27 1 61 195 64 71 1 89 176 24 28 1 111 176 18 27 1 19 174 18 21 1 47 192 33 38 1 199 133 16 21 1 180 185 25 31 1 212 181 34 38 1 244 163 20 23 1 241 148 18 19 1 265 114 15 18 1 279 126 19 19 1 315 129 20 22 1 320 150 17 27 1 298 174 24 31 1 
+JPEGImages/PartB_00014.jpg 11 2 512 65 150 1 103 569 103 96 1 215 495 181 175 1 341 510 95 154 1 474 580 178 176 1 460 539 98 105 1 573 538 130 144 1 758 552 101 124 1 813 580 130 126 1 1019 576 105 118 1 777 430 59 68 1 
+JPEGImages/PartB_01491.jpg 6 170 43 25 30 1 89 151 45 42 1 178 133 42 38 1 265 113 31 31 1 288 110 23 27 1 339 110 46 38 1 
+JPEGImages/PartB_01243.jpg 7 3 227 245 338 1 84 116 138 153 1 329 121 81 100 1 443 118 121 138 1 603 119 71 109 1 813 205 207 280 1 865 152 110 149 1 
+JPEGImages/PartB_01011.jpg 56 12 637 54 59 1 59 614 43 53 1 102 645 34 50 1 140 605 46 48 1 168 640 40 46 1 203 650 35 42 1 216 676 43 54 1 241 658 34 43 1 249 633 37 38 1 288 598 38 44 1 304 674 41 50 1 297 654 29 38 1 330 646 36 41 1 358 654 36 41 1 399 673 44 50 1 404 639 32 46 1 374 635 32 33 1 440 626 34 40 1 443 667 38 41 1 497 673 39 48 1 481 644 30 37 1 509 631 25 40 1 534 614 31 34 1 526 654 45 46 1 574 667 46 50 1 561 626 38 40 1 595 614 37 45 1 615 662 33 41 1 652 662 34 48 1 649 627 39 39 1 626 628 26 35 1 686 648 46 52 1 707 628 31 38 1 735 642 33 49 1 738 617 33 34 1 767 639 41 47 1 791 672 38 50 1 816 639 36 41 1 865 637 32 35 1 865 662 34 43 1 907 654 28 47 1 852 621 28 38 1 953 658 44 51 1 986 669 30 43 1 984 631 34 40 1 1016 657 31 53 1 1036 655 49 53 1 1076 653 32 46 1 1093 641 42 48 1 1143 650 37 49 1 1182 645 39 47 1 1213 668 49 50 1 779 777 45 53 1 876 777 46 59 1 971 772 51 56 1 1086 782 63 66 1 
+JPEGImages/PartB_01203.jpg 22 1467 1364 173 211 1 1937 1833 142 181 1 2133 1752 146 166 1 2329 1779 131 158 1 2060 1972 242 265 1 2436 2002 254 250 1 2536 1883 200 219 1 2763 1841 189 234 1 2933 1752 165 208 1 2994 1841 196 219 1 3090 1729 112 150 1 3325 1756 131 146 1 3275 1945 242 269 1 3425 1841 192 254 1 3556 1822 204 203 1 3521 1729 139 143 1 3790 1810 196 219 1 3783 1749 127 161 1 4167 1214 269 265 1 4440 1845 281 319 1 4760 1875 269 243 1 4067 1918 227 250 1 
+JPEGImages/PartB_00069.jpg 34 25 89 26 23 1 165 74 33 28 1 229 72 28 30 1 12 299 51 40 1 1 342 46 60 1 18 395 60 53 1 122 327 57 52 1 127 273 42 38 1 121 244 38 34 1 123 215 30 26 1 54 225 32 35 1 77 214 29 30 1 158 202 27 29 1 179 242 41 38 1 217 198 28 31 1 254 263 35 33 1 257 235 28 31 1 226 297 43 45 1 344 326 55 64 1 345 286 44 39 1 342 248 33 37 1 297 210 22 32 1 324 162 35 25 1 371 169 28 28 1 378 234 25 30 1 461 249 41 36 1 419 174 30 27 1 498 167 31 34 1 393 222 26 32 1 533 278 39 38 1 466 280 49 51 1 553 184 39 37 1 562 61 38 42 1 553 374 47 72 1 
+JPEGImages/PartB_02267.jpg 7 11 57 39 41 1 52 60 43 57 1 57 35 41 44 1 153 122 60 71 1 166 5 45 62 1 220 3 55 75 1 257 7 76 84 1 
+JPEGImages/PartB_01653.jpg 9 1 283 74 80 1 96 276 50 49 1 89 258 32 30 1 192 233 25 28 1 222 243 58 63 1 241 227 32 31 1 304 210 36 45 1 383 204 39 43 1 407 210 50 58 1 
+JPEGImages/PartB_01586.jpg 4 1954 399 188 227 1 467 1426 230 270 1 1527 1435 277 327 1 2286 1549 354 291 1 
+JPEGImages/PartB_01718.jpg 5 8 240 69 60 1 56 206 59 61 1 193 207 51 60 1 319 217 61 64 1 313 82 39 47 1 
+JPEGImages/PartB_01327.jpg 17 268 128 90 103 1 299 482 127 121 1 477 421 101 111 1 591 423 98 92 1 695 421 87 86 1 781 414 86 83 1 842 399 74 79 1 909 374 76 76 1 996 332 65 67 1 1031 325 50 57 1 480 648 209 229 1 690 613 200 193 1 866 531 174 163 1 873 459 152 123 1 971 429 113 106 1 1044 417 78 107 1 1156 405 122 102 1 
+JPEGImages/PartB_02224.jpg 89 5 273 11 12 1 15 268 9 13 1 9 264 7 7 1 26 278 14 14 1 26 268 10 11 1 29 265 10 8 1 44 264 9 10 1 39 273 11 12 1 25 304 29 23 1 64 273 13 12 1 64 266 9 8 1 70 261 8 10 1 86 268 11 13 1 86 265 8 9 1 104 265 11 11 1 110 272 12 9 1 118 262 11 11 1 125 266 11 10 1 142 266 9 8 1 116 276 16 16 1 127 285 19 20 1 124 303 24 22 1 105 315 41 42 1 97 293 25 25 1 93 282 18 16 1 151 282 16 14 1 157 272 13 12 1 171 275 9 10 1 169 282 15 18 1 171 297 26 26 1 193 284 16 22 1 211 284 16 17 1 227 282 15 17 1 211 274 12 11 1 213 266 11 9 1 222 270 12 10 1 234 271 10 13 1 238 268 9 9 1 190 276 12 9 1 193 272 11 9 1 247 299 28 32 1 257 294 20 23 1 262 281 16 16 1 268 274 12 13 1 286 265 10 8 1 291 271 10 10 1 299 276 13 16 1 304 280 14 19 1 311 289 22 21 1 330 272 13 13 1 312 270 6 8 1 343 267 8 7 1 365 266 12 8 1 362 285 24 24 1 347 319 49 55 1 416 291 32 37 1 405 280 21 18 1 396 278 13 19 1 377 275 12 13 1 356 272 10 11 1 377 266 10 11 1 400 265 12 11 1 404 270 9 10 1 411 271 11 10 1 434 273 14 12 1 432 282 14 17 1 449 271 13 16 1 444 266 9 9 1 462 264 9 10 1 467 265 7 8 1 484 273 14 16 1 484 282 24 28 1 497 247 11 14 1 512 275 11 11 1 520 266 10 12 1 515 262 7 8 1 534 275 13 10 1 540 285 33 40 1 526 283 14 18 1 569 299 29 26 1 582 282 15 13 1 557 267 14 15 1 573 268 12 13 1 588 272 11 13 1 542 264 9 10 1 531 261 10 14 1 555 263 8 7 1 277 228 6 6 1 314 229 6 7 1 
+JPEGImages/PartB_00000.jpg 26 19 193 24 29 1 72 206 28 28 1 95 219 34 34 1 134 212 40 39 1 134 192 18 20 1 183 187 17 18 1 207 218 57 74 1 262 202 20 25 1 277 213 30 27 1 271 191 16 17 1 323 208 26 33 1 327 201 22 28 1 307 261 92 88 1 370 208 22 25 1 376 200 15 19 1 406 192 16 21 1 419 209 20 30 1 428 204 18 23 1 449 197 15 20 1 477 211 18 22 1 484 203 16 20 1 498 207 21 22 1 520 213 16 21 1 550 225 50 51 1 563 207 23 26 1 532 290 68 115 1 
+JPEGImages/PartB_01679.jpg 5 97 186 63 73 1 79 178 48 43 1 203 188 42 48 1 247 105 33 36 1 341 169 43 41 1 
+JPEGImages/PartB_00551.jpg 7 2 233 190 239 1 73 148 189 238 1 228 221 160 208 1 316 117 163 191 1 511 128 197 252 1 758 173 140 215 1 978 157 209 295 1 
+JPEGImages/PartB_02173.jpg 22 5 339 75 100 1 48 252 33 45 1 87 239 29 30 1 116 249 39 40 1 162 259 33 40 1 171 234 23 26 1 192 228 19 22 1 203 210 16 17 1 211 239 18 18 1 220 235 21 27 1 231 245 30 34 1 212 277 59 74 1 266 265 35 44 1 271 240 27 26 1 294 251 33 37 1 323 238 19 27 1 356 248 40 48 1 317 284 60 65 1 254 245 19 20 1 540 236 23 26 1 555 244 29 29 1 581 233 19 27 1 
+JPEGImages/PartB_00258.jpg 13 104 199 18 25 1 150 203 15 17 1 183 194 16 19 1 212 193 17 22 1 247 196 18 23 1 282 200 16 20 1 311 202 16 18 1 338 198 18 25 1 367 199 17 20 1 397 199 17 20 1 431 202 20 22 1 458 205 18 22 1 490 200 17 19 1 
+JPEGImages/PartB_00861.jpg 6 139 205 144 162 1 291 176 92 102 1 415 159 125 143 1 553 234 163 185 1 718 159 133 140 1 888 213 148 167 1 
+JPEGImages/PartB_01880.jpg 6 92 199 97 95 1 122 285 48 55 1 202 222 49 68 1 241 210 66 83 1 317 236 79 99 1 443 271 100 108 1 
+JPEGImages/PartB_01476.jpg 17 8 153 47 53 1 64 135 51 53 1 145 130 38 44 1 187 135 37 39 1 188 161 90 103 1 283 156 70 81 1 292 132 37 39 1 261 115 23 32 1 229 131 35 40 1 218 115 28 31 1 308 115 32 33 1 341 121 22 30 1 390 107 35 37 1 390 142 62 67 1 399 178 121 125 1 522 155 77 93 1 522 132 37 42 1 
+JPEGImages/PartB_01056.jpg 11 317 94 15 17 1 355 156 19 19 1 421 150 20 20 1 468 149 23 26 1 525 152 24 29 1 301 142 20 18 1 235 149 21 21 1 196 158 22 18 1 172 158 18 17 1 146 144 17 15 1 45 150 19 19 1 
+JPEGImages/PartB_00636.jpg 8 676 401 179 207 1 1637 804 244 234 1 2318 783 187 230 1 2000 864 302 335 1 1511 1018 282 327 1 504 1104 377 369 1 320 1045 254 314 1 1460 1532 544 669 1 
+JPEGImages/PartB_00817.jpg 9 1 91 81 142 1 91 93 80 117 1 97 46 69 62 1 186 54 52 73 1 217 146 70 87 1 294 163 56 92 1 392 98 45 64 1 390 170 61 80 1 438 119 42 66 1 
+JPEGImages/PartB_00851.jpg 5 51 2 476 300 1 813 4 516 508 1 945 476 312 430 1 1733 160 422 508 1 1983 134 686 782 1 
+JPEGImages/PartB_01612.jpg 32 44 311 46 45 1 54 345 56 57 1 94 321 32 33 1 113 332 73 80 1 160 304 46 48 1 183 337 39 42 1 209 351 48 49 1 219 304 38 39 1 262 295 32 35 1 263 388 63 70 1 330 317 43 52 1 360 310 34 42 1 360 340 29 32 1 358 366 60 54 1 298 515 132 144 1 439 361 37 32 1 426 328 37 34 1 532 345 38 32 1 531 329 27 29 1 594 338 33 35 1 617 328 31 31 1 519 415 85 100 1 640 399 53 62 1 670 380 47 51 1 681 316 46 42 1 735 343 41 42 1 765 345 29 33 1 797 246 31 34 1 804 478 99 99 1 841 412 63 59 1 826 382 52 53 1 843 361 38 39 1 
+JPEGImages/PartB_01258.jpg 13 147 330 82 97 1 138 266 41 63 1 4 285 61 64 1 212 252 58 77 1 377 441 124 144 1 406 312 64 93 1 438 252 42 56 1 570 261 51 64 1 623 320 87 131 1 708 263 38 57 1 732 263 58 76 1 865 259 52 76 1 1102 326 91 107 1 
+JPEGImages/PartB_02327.jpg 17 152 218 64 68 1 213 193 42 46 1 265 124 27 35 1 330 181 46 44 1 335 238 80 90 1 330 159 36 34 1 413 165 27 33 1 427 178 34 39 1 463 156 17 25 1 501 161 22 27 1 489 189 47 49 1 544 177 36 41 1 532 172 24 24 1 575 174 24 28 1 559 166 15 20 1 16 202 49 53 1 3 301 95 99 1 
+JPEGImages/PartB_02401.jpg 78 22 217 41 42 1 18 198 25 26 1 26 185 20 21 1 16 181 13 14 1 7 182 12 12 1 66 187 15 17 1 81 199 21 23 1 126 210 30 31 1 142 229 52 55 1 249 206 40 40 1 206 195 26 30 1 179 191 22 18 1 133 189 22 22 1 109 184 19 19 1 98 179 14 16 1 89 175 12 15 1 115 164 11 13 1 144 176 11 14 1 164 182 12 13 1 151 193 12 16 1 125 178 10 13 1 114 175 9 11 1 138 173 9 11 1 154 175 9 11 1 161 174 9 9 1 172 171 8 11 1 185 174 9 14 1 192 179 10 14 1 195 176 9 8 1 209 181 13 15 1 223 192 18 19 1 244 193 22 24 1 290 198 20 23 1 278 184 12 14 1 265 182 15 17 1 264 174 11 10 1 245 177 12 13 1 237 184 10 11 1 228 179 12 15 1 219 172 11 12 1 279 174 8 11 1 286 174 9 10 1 297 179 13 16 1 323 186 17 19 1 333 195 20 24 1 336 218 35 38 1 317 177 8 9 1 329 176 8 8 1 335 178 8 10 1 341 182 11 13 1 344 176 10 11 1 355 182 12 14 1 357 175 8 11 1 372 179 10 13 1 373 187 18 19 1 393 208 27 27 1 440 197 28 28 1 476 194 15 18 1 491 187 13 14 1 498 178 9 13 1 506 181 10 12 1 519 178 14 15 1 530 182 10 12 1 542 179 9 13 1 552 180 11 13 1 565 181 11 12 1 432 188 12 15 1 408 179 13 18 1 391 184 12 18 1 389 174 10 12 1 401 178 9 12 1 371 173 7 9 1 417 174 10 13 1 428 179 8 7 1 437 175 8 12 1 447 176 9 11 1 462 184 9 13 1 469 179 10 13 1 
+JPEGImages/PartB_01443.jpg 46 172 714 40 49 1 139 830 96 89 1 260 713 26 38 1 328 699 34 41 1 367 783 42 38 1 305 820 50 52 1 321 876 70 76 1 430 860 53 53 1 410 798 35 39 1 430 728 32 34 1 506 698 26 37 1 578 726 41 28 1 533 778 40 37 1 496 800 35 39 1 547 845 50 58 1 584 833 39 37 1 563 809 37 40 1 649 784 34 44 1 371 1002 94 92 1 597 972 58 72 1 688 910 64 74 1 708 823 37 50 1 741 834 50 50 1 755 796 37 44 1 825 802 36 35 1 815 703 27 37 1 868 875 42 59 1 920 899 52 64 1 965 855 47 50 1 942 824 40 46 1 915 720 27 30 1 960 699 31 31 1 1023 696 38 35 1 1071 720 26 32 1 1068 762 37 41 1 1107 772 24 32 1 1123 784 26 34 1 1121 821 41 47 1 1123 879 45 51 1 1092 934 83 94 1 1199 924 59 73 1 1281 837 38 47 1 1293 804 33 38 1 1357 833 47 46 1 1346 805 38 45 1 1477 841 70 73 1 
+JPEGImages/PartB_01136.jpg 5 237 1580 508 503 1 1687 508 428 503 1 2979 1647 352 539 1 3162 1702 778 947 1 3765 1341 219 532 1 
+JPEGImages/PartB_00416.jpg 4 243 125 37 44 1 256 250 49 51 1 126 244 51 56 1 1 253 49 58 1 
+JPEGImages/PartB_00999.jpg 5 84 205 138 123 1 380 295 54 58 1 565 103 238 263 1 994 235 94 122 1 1147 285 51 88 1 
+JPEGImages/PartB_01080.jpg 7 39 58 124 138 1 130 47 51 68 1 177 92 25 40 1 195 85 34 44 1 384 82 42 43 1 469 61 52 70 1 535 30 78 86 1 
+JPEGImages/PartB_00249.jpg 8 50 245 42 44 1 89 215 65 65 1 188 249 32 46 1 237 209 54 60 1 300 232 35 45 1 376 231 24 27 1 429 239 42 53 1 502 206 50 39 1 
+JPEGImages/PartB_00339.jpg 23 88 145 56 60 1 133 119 29 36 1 163 128 38 45 1 199 189 71 84 1 230 139 46 56 1 271 107 23 32 1 291 96 21 27 1 302 88 22 28 1 337 102 24 27 1 332 131 31 36 1 394 114 23 28 1 420 117 28 35 1 244 90 21 26 1 227 101 27 31 1 216 93 17 22 1 192 92 21 24 1 150 54 15 18 1 88 103 26 29 1 71 98 26 27 1 54 77 17 15 1 31 92 20 21 1 1 107 26 35 1 2 217 48 103 1 
+JPEGImages/PartB_00405.jpg 17 88 219 77 63 1 150 208 51 52 1 25 247 63 66 1 284 237 75 72 1 373 193 78 89 1 387 129 46 51 1 487 138 44 51 1 541 208 51 47 1 511 185 39 50 1 589 219 67 80 1 588 177 66 60 1 628 119 53 60 1 702 155 52 53 1 781 193 53 63 1 843 203 78 88 1 1114 97 69 97 1 776 158 41 47 1 
+JPEGImages/PartB_01062.jpg 13 27 199 29 28 1 39 172 21 23 1 91 143 26 25 1 168 158 22 25 1 171 195 27 30 1 194 131 20 22 1 239 157 26 25 1 294 138 22 23 1 335 177 23 25 1 308 204 31 29 1 385 141 32 31 1 490 200 35 36 1 510 152 25 32 1 
+JPEGImages/PartB_01852.jpg 27 7 260 39 47 1 55 251 54 52 1 166 237 51 55 1 193 267 89 100 1 250 229 40 42 1 267 185 29 32 1 301 236 30 39 1 322 235 23 27 1 308 223 20 23 1 329 219 18 22 1 367 221 17 24 1 370 244 64 75 1 429 245 48 65 1 393 229 29 31 1 412 214 20 22 1 423 218 37 31 1 450 210 16 22 1 457 228 23 30 1 478 221 36 43 1 505 221 36 39 1 535 215 27 37 1 549 205 21 29 1 560 217 17 25 1 567 203 24 35 1 616 216 31 41 1 651 216 49 78 1 568 248 50 56 1 
+JPEGImages/PartB_02070.jpg 8 1 323 86 92 1 81 249 64 71 1 207 264 76 87 1 316 223 66 81 1 453 41 55 73 1 478 241 84 100 1 554 227 63 81 1 404 317 99 105 1 
+JPEGImages/PartB_00194.jpg 46 128 164 31 36 1 144 139 23 25 1 162 132 21 21 1 155 115 14 18 1 118 105 11 15 1 108 123 18 17 1 100 109 14 15 1 92 120 18 24 1 60 124 24 23 1 59 110 15 16 1 13 114 18 18 1 16 98 12 15 1 1 86 12 15 1 12 79 11 10 1 24 89 11 10 1 52 71 10 11 1 68 98 13 15 1 52 100 13 15 1 87 74 7 10 1 81 83 11 12 1 128 88 9 12 1 135 77 8 12 1 157 86 10 12 1 147 95 10 14 1 164 78 8 8 1 175 83 8 9 1 179 96 10 12 1 189 103 11 12 1 200 103 10 11 1 199 85 9 12 1 201 79 10 9 1 186 83 9 9 1 221 91 10 11 1 236 90 10 11 1 250 101 9 11 1 250 94 8 10 1 269 100 11 11 1 281 116 10 13 1 284 107 10 11 1 301 109 9 13 1 353 96 20 23 1 251 172 30 33 1 269 203 31 45 1 262 119 11 12 1 239 111 11 13 1 1 156 30 40 1 
+JPEGImages/PartB_01637.jpg 16 125 54 55 65 1 94 74 33 39 1 88 56 24 29 1 111 61 23 28 1 180 66 35 39 1 211 108 68 90 1 254 58 39 50 1 293 70 34 42 1 357 73 26 37 1 380 72 33 45 1 421 72 19 26 1 446 67 16 22 1 450 83 17 20 1 369 113 58 77 1 479 89 49 66 1 548 98 32 41 1 
+JPEGImages/PartB_00632.jpg 8 122 285 125 99 1 223 229 72 62 1 245 141 17 21 1 269 151 22 23 1 313 198 50 55 1 414 295 115 111 1 442 218 87 77 1 543 236 93 79 1 
+JPEGImages/PartB_01703.jpg 33 64 251 27 24 1 70 269 29 30 1 94 290 54 52 1 185 259 28 26 1 174 254 22 27 1 217 276 34 35 1 257 285 38 40 1 300 299 49 52 1 282 250 22 27 1 298 258 22 25 1 326 251 24 29 1 365 252 19 27 1 380 222 19 22 1 384 253 19 24 1 398 264 30 34 1 436 252 21 27 1 463 264 28 37 1 466 248 15 21 1 487 253 22 27 1 513 238 24 28 1 505 271 42 47 1 524 263 26 36 1 557 250 18 24 1 576 264 28 36 1 585 250 19 24 1 598 244 18 21 1 626 280 31 34 1 623 265 19 24 1 632 251 16 19 1 647 242 13 17 1 668 258 28 33 1 682 246 15 24 1 698 260 19 27 1 
+JPEGImages/PartB_00771.jpg 8 75 488 227 232 1 246 471 144 158 1 416 412 101 108 1 696 120 72 95 1 747 399 93 127 1 905 559 173 213 1 916 475 147 137 1 1113 362 85 127 1 
+JPEGImages/PartB_01910.jpg 11 26 178 34 34 1 8 168 26 29 1 1 155 13 19 1 50 150 15 17 1 83 147 13 17 1 77 172 38 41 1 156 187 50 42 1 141 149 22 24 1 151 119 13 15 1 198 155 26 31 1 257 153 26 33 1 
+JPEGImages/PartB_00233.jpg 18 1 350 36 46 1 79 321 37 46 1 114 330 35 40 1 111 416 52 51 1 191 351 34 48 1 218 356 26 39 1 270 334 30 41 1 278 387 35 45 1 317 360 34 37 1 392 411 37 46 1 402 312 30 40 1 434 330 29 35 1 475 407 38 41 1 516 312 34 42 1 553 304 29 37 1 623 344 41 47 1 676 345 36 39 1 739 323 39 39 1 
+JPEGImages/PartB_01157.jpg 33 945 1473 379 373 1 902 1136 269 267 1 573 1056 217 231 1 471 591 188 182 1 800 793 73 100 1 743 777 75 98 1 432 885 60 86 1 398 858 85 80 1 390 818 83 61 1 894 826 98 130 1 983 875 113 120 1 1181 797 68 90 1 1126 754 78 100 1 1288 781 67 110 1 1361 832 106 106 1 1338 948 141 151 1 1547 963 189 185 1 1706 1069 206 220 1 2155 1120 302 285 1 2469 989 121 217 1 2473 863 106 132 1 2243 834 104 114 1 2157 793 77 94 1 2151 679 61 71 1 1988 756 59 82 1 1775 814 92 95 1 1800 883 143 135 1 1671 763 94 110 1 1734 760 51 62 1 1845 740 65 72 1 1883 695 55 65 1 1516 754 67 92 1 1430 754 58 74 1 
+JPEGImages/PartB_01462.jpg 56 22 525 70 56 1 123 370 37 32 1 248 415 46 40 1 265 378 41 41 1 440 441 50 46 1 324 320 26 24 1 346 326 24 24 1 395 338 26 30 1 388 332 19 17 1 419 317 17 19 1 443 309 19 23 1 455 319 24 20 1 455 342 30 29 1 500 302 23 26 1 462 259 18 19 1 521 297 24 26 1 546 318 20 22 1 558 333 26 30 1 570 325 26 23 1 562 297 18 21 1 577 306 23 17 1 605 291 16 22 1 621 309 20 29 1 594 360 27 25 1 643 360 34 37 1 719 358 40 40 1 700 344 23 31 1 660 309 25 24 1 635 281 15 19 1 689 290 17 19 1 698 277 20 18 1 709 296 23 23 1 718 315 23 23 1 732 303 19 22 1 732 278 19 21 1 759 276 22 20 1 766 305 19 23 1 786 319 32 25 1 756 334 26 38 1 782 356 36 30 1 846 316 25 27 1 818 290 21 24 1 798 287 17 22 1 855 277 24 27 1 895 275 22 26 1 908 295 19 20 1 935 321 30 26 1 976 319 28 31 1 941 308 20 25 1 876 308 24 27 1 794 393 47 49 1 744 498 76 73 1 973 522 84 92 1 997 420 60 61 1 1114 367 46 42 1 1165 439 35 58 1 
+JPEGImages/PartB_02262.jpg 9 14 318 96 85 1 59 258 78 78 1 194 297 81 92 1 248 257 60 55 1 303 241 62 60 1 391 229 58 63 1 404 279 84 89 1 485 219 53 66 1 442 99 30 31 1 
+JPEGImages/PartB_01439.jpg 5 55 234 34 33 1 74 255 40 43 1 104 302 52 61 1 252 264 45 55 1 291 159 40 43 1 
+JPEGImages/PartB_00447.jpg 8 61 47 128 136 1 187 159 30 29 1 258 151 33 33 1 250 110 19 29 1 308 135 60 73 1 358 119 109 86 1 457 140 46 56 1 498 133 38 52 1 
+JPEGImages/PartB_00094.jpg 9 1 168 19 55 1 26 196 46 45 1 66 163 33 40 1 100 163 49 51 1 198 102 58 59 1 238 193 51 58 1 279 177 46 57 1 331 197 61 62 1 382 188 51 62 1 
+JPEGImages/PartB_01627.jpg 4 565 337 277 313 1 1131 577 165 198 1 1602 685 471 540 1 1590 1206 708 703 1 
+JPEGImages/PartB_01363.jpg 24 77 406 87 86 1 152 406 69 79 1 139 372 53 54 1 170 346 36 40 1 83 370 54 43 1 54 345 43 50 1 65 331 36 33 1 214 349 45 56 1 253 354 33 43 1 235 326 36 37 1 307 371 59 76 1 344 373 45 61 1 309 328 32 40 1 390 359 40 44 1 393 330 35 32 1 432 302 115 144 1 414 552 140 174 1 616 183 114 142 1 811 326 33 41 1 559 376 47 61 1 609 384 34 56 1 591 349 44 47 1 630 325 26 39 1 560 322 29 42 1 
+JPEGImages/PartB_00607.jpg 22 1 480 88 129 1 208 395 76 79 1 312 165 65 68 1 214 562 124 127 1 311 465 86 98 1 330 411 95 109 1 561 215 60 47 1 549 499 112 136 1 549 444 92 107 1 750 442 105 122 1 835 228 65 70 1 873 473 98 149 1 980 424 114 115 1 990 557 148 189 1 1233 618 203 248 1 1158 463 100 129 1 1247 427 83 94 1 1276 373 55 67 1 1356 440 94 125 1 1408 415 88 94 1 1401 376 76 82 1 1541 410 59 100 1 
+JPEGImages/PartB_02382.jpg 16 15 133 43 43 1 47 102 37 41 1 96 65 27 31 1 150 57 18 20 1 171 121 38 44 1 183 111 30 35 1 189 90 31 35 1 243 82 23 28 1 233 138 57 67 1 266 185 97 85 1 278 83 17 20 1 298 77 15 19 1 311 87 16 20 1 327 80 23 26 1 330 97 24 23 1 364 97 36 37 1 
+JPEGImages/PartB_00469.jpg 5 138 128 64 84 1 195 119 57 60 1 359 70 39 51 1 391 112 48 62 1 454 139 65 76 1 
+JPEGImages/PartB_01635.jpg 13 21 248 56 58 1 9 223 37 36 1 55 208 32 33 1 98 220 39 47 1 120 206 31 33 1 280 209 24 31 1 303 215 32 37 1 344 214 26 27 1 364 164 21 25 1 373 215 30 32 1 358 240 54 62 1 472 241 57 57 1 479 226 24 23 1 
+JPEGImages/PartB_01456.jpg 31 942 1290 137 161 1 167 1829 224 225 1 636 2260 682 706 1 1042 1857 367 424 1 1018 1775 237 197 1 1136 1714 185 209 1 1403 1666 121 154 1 1467 1748 148 178 1 1576 1796 209 230 1 1761 1908 303 337 1 2223 2278 425 512 1 2458 1929 321 340 1 2297 1711 191 194 1 2051 1845 179 169 1 1927 1796 155 118 1 1752 1657 157 148 1 2109 1620 149 176 1 2479 1675 130 133 1 2767 1684 178 191 1 2679 1820 200 194 1 3185 1878 282 315 1 3661 1902 366 473 1 3464 1790 257 300 1 3500 1720 167 146 1 3445 1648 134 124 1 3233 1817 179 167 1 3112 1739 161 145 1 3158 1663 130 130 1 3685 1642 151 154 1 3791 1702 167 164 1 3909 1596 123 182 1 
+JPEGImages/PartB_01504.jpg 18 29 1083 208 214 1 211 1134 143 167 1 268 1071 104 134 1 354 1077 137 153 1 568 977 310 369 1 860 1071 102 141 1 1007 1071 228 251 1 1095 681 126 167 1 1360 1038 100 143 1 1415 1081 169 202 1 1515 1040 82 84 1 1427 1028 59 98 1 1619 1109 143 172 1 1717 1048 100 131 1 1786 1014 92 132 1 1927 1012 66 83 1 2052 561 308 367 1 1933 1001 388 498 1 
+JPEGImages/PartB_00549.jpg 6 140 4 153 175 1 340 74 76 104 1 317 93 46 59 1 393 59 52 76 1 439 85 27 49 1 461 78 24 44 1 
+JPEGImages/PartB_01823.jpg 7 7 238 75 68 1 11 221 56 46 1 124 230 65 58 1 100 371 104 79 1 311 269 85 85 1 366 223 65 65 1 415 123 32 41 1 
+JPEGImages/PartB_01641.jpg 21 29 138 21 22 1 46 146 24 22 1 78 150 22 25 1 98 156 27 26 1 155 148 19 20 1 175 147 20 22 1 202 138 13 15 1 218 148 19 20 1 215 166 37 39 1 264 148 16 18 1 277 147 17 21 1 307 136 18 21 1 379 112 22 23 1 281 170 42 47 1 338 158 25 32 1 388 159 32 35 1 415 147 24 23 1 447 155 24 28 1 466 174 38 46 1 502 185 70 72 1 526 154 36 38 1 
+JPEGImages/PartB_01808.jpg 13 43 206 64 54 1 41 249 30 35 1 62 156 42 42 1 117 184 73 77 1 119 164 34 38 1 139 146 44 45 1 176 124 59 57 1 225 88 53 60 1 264 72 62 67 1 223 179 86 85 1 380 192 30 29 1 443 114 112 127 1 427 183 23 33 1 
+JPEGImages/PartB_00099.jpg 39 361 80 103 104 1 355 50 48 41 1 288 59 62 65 1 265 74 33 48 1 155 89 34 41 1 165 56 30 32 1 200 56 20 25 1 219 65 33 41 1 262 51 25 32 1 242 52 23 24 1 297 31 15 14 1 289 20 13 14 1 141 21 13 13 1 135 4 13 12 1 158 2 9 11 1 164 13 10 17 1 175 27 12 16 1 142 57 17 23 1 153 36 15 13 1 201 28 14 15 1 219 23 11 13 1 238 28 12 14 1 273 11 12 12 1 275 1 12 7 1 259 3 10 15 1 240 9 11 12 1 247 3 9 9 1 236 1 9 9 1 223 5 11 13 1 204 15 9 13 1 200 12 8 11 1 185 2 10 10 1 176 11 8 11 1 166 1 10 7 1 215 1 11 10 1 196 42 18 18 1 227 38 12 15 1 249 41 14 15 1 252 33 17 14 1 
+JPEGImages/PartB_01376.jpg 7 1 77 23 33 1 11 56 20 21 1 33 78 39 34 1 83 65 25 29 1 96 60 17 22 1 199 63 28 26 1 230 22 51 60 1 
+JPEGImages/PartB_01084.jpg 51 175 102 9 11 1 358 97 11 16 1 209 182 47 57 1 282 209 91 115 1 470 174 62 96 1 396 154 39 36 1 446 143 34 34 1 495 127 28 30 1 477 124 17 22 1 450 126 18 18 1 432 121 19 21 1 422 137 21 22 1 402 133 18 21 1 410 125 14 13 1 388 121 16 19 1 360 122 14 13 1 336 138 19 25 1 352 136 16 19 1 333 132 14 14 1 339 124 10 11 1 324 123 10 13 1 307 154 37 44 1 287 153 26 32 1 266 137 17 20 1 277 133 14 11 1 259 127 11 12 1 267 122 8 12 1 300 123 8 11 1 218 134 19 26 1 241 130 11 11 1 208 131 12 12 1 201 127 11 9 1 177 134 15 17 1 151 142 23 22 1 154 131 15 17 1 131 133 13 15 1 142 131 11 12 1 124 139 15 21 1 120 156 26 31 1 116 131 12 13 1 107 134 14 18 1 97 138 10 15 1 65 139 24 27 1 77 132 14 14 1 54 141 20 16 1 37 140 17 16 1 43 129 14 15 1 2 176 42 58 1 11 164 29 22 1 19 145 17 18 1 9 137 15 15 1 
+JPEGImages/PartB_00977.jpg 6 194 163 36 41 1 102 291 94 92 1 193 261 58 61 1 379 290 84 97 1 346 264 63 64 1 570 281 30 92 1 
+JPEGImages/PartB_01517.jpg 25 1 91 20 19 1 39 92 21 18 1 71 91 19 16 1 65 79 13 12 1 93 80 12 14 1 104 92 18 19 1 121 86 11 12 1 118 77 11 10 1 124 74 9 10 1 138 72 9 12 1 146 79 10 12 1 142 90 17 19 1 172 83 15 16 1 168 79 9 11 1 156 74 8 9 1 158 52 7 8 1 173 68 9 10 1 183 81 11 12 1 181 75 7 8 1 187 70 8 8 1 196 74 7 10 1 178 58 7 7 1 245 76 8 11 1 248 73 7 6 1 285 99 18 18 1 
+JPEGImages/PartB_00995.jpg 34 90 344 66 85 1 165 366 51 57 1 148 344 35 38 1 177 332 25 34 1 148 331 20 20 1 169 328 18 23 1 205 328 20 23 1 247 329 20 24 1 280 335 24 32 1 265 361 49 51 1 299 344 68 85 1 307 322 20 23 1 326 332 15 22 1 343 329 14 20 1 350 327 14 17 1 361 343 24 32 1 390 331 17 23 1 405 324 14 20 1 422 326 18 26 1 435 347 31 37 1 448 360 50 67 1 456 330 21 24 1 475 321 12 26 1 490 332 22 29 1 507 353 35 54 1 536 331 32 34 1 531 326 17 23 1 597 327 22 34 1 633 337 53 61 1 726 326 29 34 1 797 323 37 44 1 817 360 47 56 1 890 345 38 50 1 912 340 32 48 1 
+JPEGImages/PartB_02171.jpg 34 274 251 46 50 1 369 232 30 45 1 342 221 26 28 1 374 203 17 16 1 399 208 17 22 1 413 203 16 17 1 446 212 19 18 1 451 235 28 31 1 501 222 26 29 1 524 215 17 20 1 545 210 15 19 1 325 205 17 22 1 328 196 14 15 1 313 195 15 17 1 270 212 19 20 1 269 200 14 16 1 285 201 14 15 1 220 195 17 18 1 246 199 8 14 1 207 203 17 20 1 185 211 22 28 1 139 227 31 40 1 127 212 26 23 1 124 196 17 15 1 149 187 15 19 1 169 200 15 20 1 184 168 8 13 1 103 186 13 17 1 91 186 15 19 1 69 196 22 21 1 79 216 26 35 1 29 208 28 28 1 44 191 17 21 1 31 161 15 13 1 
+JPEGImages/PartB_01803.jpg 7 70 210 120 145 1 85 128 110 111 1 235 12 108 130 1 259 236 116 127 1 354 146 115 143 1 395 118 94 99 1 471 131 129 135 1 
+JPEGImages/PartB_00666.jpg 6 222 64 214 265 1 141 164 98 118 1 575 124 124 144 1 711 104 109 135 1 857 149 107 111 1 1078 168 99 109 1 
+JPEGImages/PartB_00217.jpg 22 14 77 44 48 1 109 88 39 47 1 91 64 27 32 1 121 53 19 23 1 95 50 15 17 1 103 42 14 17 1 73 46 17 22 1 55 51 22 30 1 29 59 32 28 1 153 4 15 18 1 179 42 14 17 1 189 41 15 21 1 198 42 18 23 1 215 43 14 19 1 227 47 22 27 1 240 40 20 24 1 268 59 27 31 1 279 47 15 20 1 265 39 12 18 1 297 40 14 20 1 318 41 14 18 1 338 46 12 17 1 
+JPEGImages/PartB_02107.jpg 26 19 207 23 20 1 26 190 21 19 1 12 200 18 13 1 59 200 29 33 1 62 187 17 20 1 94 185 18 21 1 124 188 20 25 1 32 263 69 47 1 166 192 19 18 1 203 186 25 23 1 231 183 18 19 1 253 154 13 20 1 207 206 29 36 1 255 194 20 28 1 256 222 31 34 1 261 256 59 57 1 290 197 26 26 1 322 186 17 20 1 330 192 20 20 1 350 192 21 23 1 367 206 28 39 1 393 188 14 18 1 409 189 17 22 1 430 198 16 18 1 449 197 21 28 1 487 190 17 19 1 
+JPEGImages/PartB_00279.jpg 10 496 93 101 111 1 380 97 85 81 1 262 90 53 53 1 246 77 43 49 1 214 59 45 48 1 146 55 33 32 1 91 53 28 31 1 50 72 45 44 1 65 61 34 30 1 6 74 42 42 1 
+JPEGImages/PartB_01054.jpg 116 2 92 15 17 1 2 76 9 9 1 10 82 10 9 1 17 87 12 10 1 28 96 18 24 1 43 86 14 16 1 23 77 11 11 1 13 69 11 11 1 42 75 10 10 1 55 79 12 10 1 49 69 10 8 1 60 88 16 15 1 55 106 27 25 1 83 101 25 29 1 99 107 18 24 1 108 102 21 27 1 120 108 29 33 1 162 105 33 35 1 194 113 35 42 1 252 113 45 49 1 292 92 27 31 1 277 80 21 24 1 255 78 16 24 1 222 82 26 27 1 212 86 23 26 1 191 80 17 22 1 173 83 15 16 1 153 86 21 24 1 170 72 12 13 1 186 69 13 11 1 159 72 10 10 1 151 70 10 13 1 134 76 13 11 1 131 69 12 10 1 120 74 13 15 1 114 87 13 14 1 81 83 12 17 1 99 87 12 13 1 102 81 14 12 1 102 67 12 13 1 91 77 12 12 1 68 81 12 14 1 68 73 12 12 1 71 67 13 10 1 138 50 13 14 1 152 41 13 12 1 178 48 14 14 1 210 65 10 13 1 216 60 11 14 1 210 50 10 11 1 217 46 12 14 1 233 40 13 15 1 245 48 11 12 1 258 48 12 13 1 277 37 12 15 1 351 147 49 55 1 430 129 56 61 1 482 124 48 76 1 505 112 94 108 1 480 100 36 42 1 434 93 34 39 1 372 93 32 33 1 327 95 30 33 1 337 69 24 25 1 320 80 21 20 1 310 76 13 18 1 295 66 13 16 1 305 64 11 14 1 295 47 9 11 1 313 46 10 14 1 324 40 11 12 1 324 61 14 16 1 317 65 12 13 1 338 39 13 18 1 350 36 9 10 1 349 16 9 10 1 361 37 13 16 1 354 49 11 11 1 366 64 22 22 1 379 78 25 24 1 404 74 27 30 1 396 59 20 22 1 375 37 13 19 1 405 32 14 16 1 392 19 9 12 1 430 30 15 18 1 411 20 9 9 1 419 16 9 10 1 434 13 10 12 1 414 50 17 22 1 428 62 17 21 1 452 60 24 28 1 463 77 25 30 1 512 77 20 28 1 529 58 30 41 1 498 51 25 28 1 487 56 16 21 1 479 31 20 22 1 465 25 14 18 1 459 39 11 13 1 442 21 11 13 1 456 11 10 13 1 505 30 9 10 1 512 38 12 15 1 524 25 16 20 1 516 30 8 11 1 546 50 18 17 1 561 51 16 22 1 586 56 14 19 1 588 33 12 19 1 572 23 11 16 1 564 38 13 15 1 560 27 11 12 1 553 30 8 10 1 546 39 13 13 1 540 24 10 13 1 
+JPEGImages/PartB_01600.jpg 18 205 1103 212 184 1 5 1014 108 151 1 701 1075 85 79 1 876 862 86 109 1 893 1087 122 157 1 780 1240 498 500 1 1127 1075 112 145 1 1197 1077 87 104 1 1272 1091 106 114 1 1621 1171 131 145 1 1827 1124 112 128 1 1811 1118 73 79 1 1517 967 65 79 1 2064 1150 71 86 1 2101 1124 69 75 1 2117 1173 133 143 1 2231 1207 302 280 1 2201 1152 124 129 1 
+JPEGImages/PartB_00745.jpg 11 71 107 29 34 1 120 118 33 35 1 179 108 32 38 1 229 114 35 36 1 101 172 38 45 1 162 229 57 73 1 225 161 39 56 1 363 151 35 42 1 266 218 55 60 1 367 209 57 63 1 466 191 54 63 1

examples/pytorch/vision/Face_Detection/data/widerface.py

@@ -7,13 +7,19 @@ import torch.utils.data as data
 import numpy as np
 import random
 import sys; sys.path.append('../')
-from utils.augmentations import preprocess
+from utils.augmentations import preprocess, preprocess_qvga
+from data.choose_config import cfg
+cfg = cfg.cfg
+import os
+
+HOME = os.environ['DATA_HOME']
+SCUT_ROOT = os.path.join(HOME, 'SCUT_HEAD_Part_B')
 
 
 class WIDERDetection(data.Dataset):
     """docstring for WIDERDetection"""
 
-    def __init__(self, list_file, mode='train', mono_mode=False):
+    def __init__(self, list_file, mode='train', mono_mode=False, is_scut=False):
         super(WIDERDetection, self).__init__()
         self.mode = mode
         self.mono_mode = mono_mode
@@ -40,7 +46,10 @@ class WIDERDetection(data.Dataset):
                 box.append([x, y, x + w, y + h])
                 label.append(c)
             if len(box) > 0:
-                self.fnames.append(line[0])
+                if is_scut==True:
+                    self.fnames.append(SCUT_ROOT + '/' + line[0])
+                else:
+                    self.fnames.append(line[0])
                 self.boxes.append(box)
                 self.labels.append(label)
 
@@ -65,8 +74,13 @@ class WIDERDetection(data.Dataset):
                 np.array(self.boxes[index]), im_width, im_height)
             label = np.array(self.labels[index])
             bbox_labels = np.hstack((label[:, np.newaxis], boxes)).tolist()
-            img, sample_labels = preprocess(
-                img, bbox_labels, self.mode, image_path)
+
+            if self.mono_mode==False:
+                img, sample_labels = preprocess(
+                    img, bbox_labels, self.mode, image_path)
+            else:
+                img, sample_labels = preprocess_qvga(
+                    img, bbox_labels, self.mode, image_path)
             sample_labels = np.array(sample_labels)
             if len(sample_labels) > 0:
                 target = np.hstack(

examples/pytorch/vision/Face_Detection/eval.py

@@ -21,6 +21,8 @@ from utils.augmentations import to_chw_bgr
 
 from importlib import import_module
 
+def str2bool(v):
+    return v.lower() in ("yes", "true", "t", "1")
 
 parser = argparse.ArgumentParser(description='face detection demo')
 parser.add_argument('--save_dir', type=str, default='results/',
@@ -29,11 +31,17 @@ parser.add_argument('--model', type=str,
                     default='weights/rpool_face_c.pth', help='trained model')
 parser.add_argument('--thresh', default=0.17, type=float,
                     help='Final confidence threshold')
+parser.add_argument('--multigpu',
+                    default=False, type=str2bool,
+                    help='Specify whether model was trained with multigpu')
 parser.add_argument('--model_arch',
                     default='RPool_Face_C', type=str,
-                    choices=['RPool_Face_C', 'RPool_Face_Quant', 'RPool_Face_QVGA_monochrome'],
+                    choices=['RPool_Face_C', 'RPool_Face_Quant', 'RPool_Face_QVGA_monochrome', 'RPool_Face_M4'],
                     help='choose architecture among rpool variants')
 parser.add_argument('--image_folder', default=None, type=str, help='folder containing images')
+parser.add_argument('--save_traces',
+                    default=False, type=str2bool,
+                    help='Specify whether to save input output traces')
 
 
 args = parser.parse_args()
@@ -49,7 +57,7 @@ else:
     torch.set_default_tensor_type('torch.FloatTensor')
 
 
-def detect(net, img_path, thresh):
+def detect(net, img_path, thresh, save_traces):
     img = Image.open(img_path)
     img = img.convert('RGB')
     img = np.array(img)
@@ -62,9 +70,14 @@ def detect(net, img_path, thresh):
         max_im_shrink = np.sqrt(
             640 * 480 / (img.shape[0] * img.shape[1]))
 
-    image = cv2.resize(img, None, None, fx=max_im_shrink,
+    if save_traces==True and os.environ['IS_QVGA_MONO'] == '1':
+        image = cv2.resize(img, (320, 240))
+    elif save_traces==True:
+        image = cv2.resize(img, (640, 480))
+    else:
+        image = cv2.resize(img, None, None, fx=max_im_shrink,
                       fy=max_im_shrink, interpolation=cv2.INTER_LINEAR)
-    # img = cv2.resize(img, (640, 640))
+
     x = to_chw_bgr(image)
     x = x.astype('float32')
     x -= cfg.img_mean
@@ -79,7 +92,7 @@ def detect(net, img_path, thresh):
     if use_cuda:
         x = x.cuda()
     t1 = time.time()
-    y = net(x)
+    y, loc, conf = net(x)
     detections = y.data
     scale = torch.Tensor([img.shape[1], img.shape[0],
                           img.shape[1], img.shape[0]])
@@ -94,9 +107,9 @@ def detect(net, img_path, thresh):
             left_up, right_bottom = (pt[0], pt[1]), (pt[2], pt[3])
             j += 1
             cv2.rectangle(img, left_up, right_bottom, (0, 0, 255), 2)
-            conf = "{:.3f}".format(score)
+            conf_score = "{:.3f}".format(score)
             point = (int(left_up[0]), int(left_up[1] - 5))
-            cv2.putText(img, conf, point, cv2.FONT_HERSHEY_COMPLEX,
+            cv2.putText(img, conf_score, point, cv2.FONT_HERSHEY_COMPLEX,
                        0.6, (0, 255, 0), 1)
 
     t2 = time.time()
@@ -104,13 +117,17 @@ def detect(net, img_path, thresh):
 
     cv2.imwrite(os.path.join(args.save_dir, os.path.basename(img_path)), img)
 
+    if save_traces == True:
+        return x, loc, conf
+
 
 if __name__ == '__main__':
 
     module = import_module('models.' + args.model_arch)
     net = module.build_s3fd('test', cfg.NUM_CLASSES)
 
-    net = torch.nn.DataParallel(net)
+    if args.multigpu == True:
+        net = torch.nn.DataParallel(net)
 
     checkpoint_dict = torch.load(args.model)
 
@@ -129,5 +146,18 @@ if __name__ == '__main__':
     img_path = args.image_folder
     img_list = [os.path.join(img_path, x)
                 for x in os.listdir(img_path)]
+    x = []
+    loc = []
+    conf = []
     for path in img_list:
-        detect(net, path, args.thresh)
+        if args.save_traces == True:
+            x_temp, loc_temp, conf_temp = detect(net, path, args.thresh, args.save_traces)
+            x.append(x_temp)
+            loc.append(loc_temp)
+            conf.append(conf_temp)
+        else:
+            detect(net, path, args.thresh, args.save_traces)
+
+    if args.save_traces == True:
+        np.save('trace_inputs.npy', torch.cat(x).cpu().detach().numpy())
+        np.save('trace_outputs.npy', torch.cat([torch.cat(conf), torch.cat(loc)], dim=1).cpu().detach().numpy())

examples/pytorch/vision/Face_Detection/layers/functions/prior_box.py

@@ -48,4 +48,4 @@ class PriorBox(object):
         output = torch.Tensor(mean).view(-1, 4)
         if self.clip:
             output.clamp_(max=1, min=0)
-        return output
+        return output

examples/pytorch/vision/Face_Detection/layers/modules/l2norm.py

@@ -24,6 +24,3 @@ class L2Norm(nn.Module):
         x = torch.div(x,norm)
         out = self.weight.unsqueeze(0).unsqueeze(2).unsqueeze(3).expand_as(x) * x
         return out
-
-
-        

examples/pytorch/vision/Face_Detection/layers/modules/multibox_loss.py

@@ -86,7 +86,7 @@ class MultiBoxLoss(nn.Module):
         pos_idx = pos.unsqueeze(pos.dim()).expand_as(loc_data)
         loc_p = loc_data[pos_idx].view(-1, 4)
         loc_t = loc_t[pos_idx].view(-1, 4)
-        loss_l = F.smooth_l1_loss(loc_p, loc_t, size_average=False)
+        loss_l = F.smooth_l1_loss(loc_p, loc_t, reduction='sum')
         # print(loc_p)
         # Compute max conf across batch for hard negative mining
         batch_conf = conf_data.view(-1, self.num_classes)
@@ -109,10 +109,10 @@ class MultiBoxLoss(nn.Module):
         conf_p = conf_data[(pos_idx + neg_idx).gt(0)
                            ].view(-1, self.num_classes)
         targets_weighted = conf_t[(pos + neg).gt(0)]
-        loss_c = F.cross_entropy(conf_p, targets_weighted, size_average=False)
+        loss_c = F.cross_entropy(conf_p, targets_weighted, reduction='sum')
 
         # Sum of losses: L(x,c,l,g) = (Lconf(x, c) + αLloc(x,l,g)) / N
         N = num_pos.data.sum() if num_pos.data.sum() > 0 else num
         loss_l /= N
         loss_c /= N
-        return loss_l, loss_c
+        return loss_l, loss_c

examples/pytorch/vision/Face_Detection/models/RPool_Face_C.py

@@ -171,7 +171,7 @@ class S3FD(nn.Module):
 
        
         if self.phase == 'test':
-            output = detect_function(
+            output = detect_function(cfg,
                 loc.view(loc.size(0), -1, 4),                   # loc preds
                 self.softmax(conf.view(conf.size(0), -1,
                                        self.num_classes)),                # conf preds
@@ -184,7 +184,7 @@ class S3FD(nn.Module):
                 conf.view(conf.size(0), -1, self.num_classes),
                 self.priors
             )
-        return output
+        return output, loc, conf
 
     def load_weights(self, base_file):
         other, ext = os.path.splitext(base_file)
@@ -304,7 +304,6 @@ class MobileNetV2(nn.Module):
 
         # building first layer
         input_channel = _make_divisible(input_channel * width_mult, round_nearest)
-        self.last_channel = _make_divisible(last_channel * max(1.0, width_mult), round_nearest)
         self.layers = []
         # building inverted residual blocks
         for t, c, n, s in inverted_residual_setting:
@@ -379,4 +378,3 @@ if __name__ == '__main__':
     net = build_s3fd('train', num_classes=2)
     inputs = Variable(torch.randn(4, 3, 640, 640))
     output = net(inputs)
-

examples/pytorch/vision/Face_Detection/models/RPool_Face_M4.py

@@ -0,0 +1,349 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT license.
+
+from __future__ import division
+from __future__ import absolute_import
+from __future__ import print_function
+
+import os
+import torch
+import torch.nn as nn
+import torch.nn.init as init
+import torch.nn.functional as F
+
+
+from layers import *
+from data.config_qvga import cfg
+import numpy as np
+
+from edgeml_pytorch.graph.rnnpool import *
+
+class S3FD(nn.Module):
+    """Single Shot Multibox Architecture
+    The network is composed of a base VGG network followed by the
+    added multibox conv layers.  Each multibox layer branches into
+        1) conv2d for class conf scores
+        2) conv2d for localization predictions
+        3) associated priorbox layer to produce default bounding
+           boxes specific to the layer's feature map size.
+    See: https://arxiv.org/pdf/1512.02325.pdf for more details.
+    Args:
+        phase: (string) Can be "test" or "train"
+        size: input image size
+        base: VGG16 layers for input, size of either 300 or 500
+        extras: extra layers that feed to multibox loc and conf layers
+        head: "multibox head" consists of loc and conf conv layers
+    """
+
+    def __init__(self, phase, base, head, num_classes):
+        super(S3FD, self).__init__()
+        self.phase = phase
+        self.num_classes = num_classes
+        '''
+        self.priorbox = PriorBox(size,cfg)
+        self.priors = Variable(self.priorbox.forward(), volatile=True)
+        '''
+        # SSD network
+        self.conv = ConvBNReLU(1, 4, stride=2)
+
+        self.unfold = nn.Unfold(kernel_size=(8,8),stride=(4,4))
+
+        self.rnn_model = RNNPool(8, 8, 16, 16, 4, 
+                            w1Sparsity=0.5, u1Sparsity=0.3, w2Sparsity=0.3, u2Sparsity=0.3)
+        
+        self.mob = nn.ModuleList(base)
+        # Layer learns to scale the l2 normalized features from conv4_3
+        self.L2Norm3_3 = L2Norm(32, 10)
+        self.L2Norm4_3 = L2Norm(32, 8)
+        self.L2Norm5_3 = L2Norm(64, 5)
+
+
+        self.loc = nn.ModuleList(head[0])
+        self.conf = nn.ModuleList(head[1])
+
+                
+        if self.phase == 'test':
+            self.softmax = nn.Softmax(dim=-1) 
+
+
+    def forward(self, x):
+        """Applies network layers and ops on input image(s) x.
+        Args:
+            x: input image or batch of images. Shape: [batch,3,300,300].
+        Return:
+            Depending on phase:
+            test:
+                Variable(tensor) of output class label predictions,
+                confidence score, and corresponding location predictions for
+                each object detected. Shape: [batch,topk,7]
+            train:
+                list of concat outputs from:
+                    1: confidence layers, Shape: [batch*num_priors,num_classes]
+                    2: localization layers, Shape: [batch,num_priors*4]
+                    3: priorbox layers, Shape: [2,num_priors*4]
+        """
+        size = x.size()[2:]
+        batch_size = x.shape[0]
+        sources = list()
+        loc = list()
+        conf = list()
+
+        x = self.conv(x)
+
+        patches = self.unfold(x)
+        patches = torch.cat(torch.unbind(patches,dim=2),dim=0)
+        patches = torch.reshape(patches,(-1,4,8,8))
+
+        output_x = int((x.shape[2]-8)/4 + 1)
+        output_y = int((x.shape[3]-8)/4 + 1)
+
+        rnnX = self.rnn_model(patches, int(batch_size)*output_x*output_y)
+
+        x = torch.stack(torch.split(rnnX, split_size_or_sections=int(batch_size), dim=0),dim=2)
+
+        x = F.fold(x, kernel_size=(1,1), output_size=(output_x,output_y))
+
+        x = F.pad(x, (0,1,0,1), mode='replicate')
+
+        for k in range(1):
+            x = self.mob[k](x)
+
+        s = self.L2Norm3_3(x)
+        sources.append(s)
+
+        # apply vgg up to fc7
+        for k in range(1, 2):
+            x = self.mob[k](x)
+
+        s = self.L2Norm4_3(x)
+        sources.append(s)
+
+        for k in range(2, 3):
+            x = self.mob[k](x)
+
+        s = self.L2Norm5_3(x)
+        sources.append(s)
+
+        for k in range(3, 4):
+            x = self.mob[k](x)
+        sources.append(x)
+
+      
+        # apply multibox head to source layers
+
+        loc_x = self.loc[0](sources[0])
+        conf_x = self.conf[0](sources[0])
+
+        max_conf, _ = torch.max(conf_x[:, 0:3, :, :], dim=1, keepdim=True)
+        conf_x = torch.cat((max_conf, conf_x[:, 3:, :, :]), dim=1)
+
+        loc.append(loc_x.permute(0, 2, 3, 1).contiguous())
+        conf.append(conf_x.permute(0, 2, 3, 1).contiguous())
+
+        for i in range(1, len(sources)):
+            x = sources[i]
+            conf.append(self.conf[i](x).permute(0, 2, 3, 1).contiguous())
+            loc.append(self.loc[i](x).permute(0, 2, 3, 1).contiguous())
+
+
+        features_maps = []
+        for i in range(len(loc)):
+            feat = []
+            feat += [loc[i].size(1), loc[i].size(2)]
+            features_maps += [feat]
+
+        self.priorbox = PriorBox(size, features_maps, cfg)
+        
+        self.priors = self.priorbox.forward()
+
+        loc = torch.cat([o.view(o.size(0), -1) for o in loc], 1)
+        conf = torch.cat([o.view(o.size(0), -1) for o in conf], 1)
+
+       
+        if self.phase == 'test':
+            output = detect_function(cfg,
+                loc.view(loc.size(0), -1, 4),                   # loc preds
+                self.softmax(conf.view(conf.size(0), -1,
+                                       self.num_classes)),      # conf preds
+                self.priors.type(type(x.data))                  # default boxes
+            )
+
+        else:
+            output = (
+                loc.view(loc.size(0), -1, 4),
+                conf.view(conf.size(0), -1, self.num_classes),
+                self.priors
+            )
+        return output, loc, conf
+
+    def load_weights(self, base_file):
+        other, ext = os.path.splitext(base_file)
+        if ext == '.pkl' or '.pth':
+            print('Loading weights into state dict...')
+            mdata = torch.load(base_file,
+                               map_location=lambda storage, loc: storage)
+            weights = mdata['weight']
+            epoch = mdata['epoch']
+            self.load_state_dict(weights)
+            print('Finished!')
+        else:
+            print('Sorry only .pth and .pkl files supported.')
+        return epoch
+
+    def xavier(self, param):
+        init.xavier_uniform(param)
+
+    def weights_init(self, m):
+        if isinstance(m, nn.Conv2d):
+            self.xavier(m.weight.data)
+            m.bias.data.zero_()
+
+
+
+
+def _make_divisible(v, divisor, min_value=None):
+    """
+    This function is taken from the original tf repo.
+    It ensures that all layers have a channel number that is divisible by 8
+    It can be seen here:
+    https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py
+    :param v:
+    :param divisor:
+    :param min_value:
+    :return:
+    """
+    if min_value is None:
+        min_value = divisor
+    new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
+    # Make sure that round down does not go down by more than 10%.
+    if new_v < 0.9 * v:
+        new_v += divisor
+    return new_v
+
+
+class ConvBNReLU(nn.Sequential):
+    def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, groups=1):
+        padding = (kernel_size - 1) // 2
+        super(ConvBNReLU, self).__init__(
+            nn.Conv2d(in_planes, out_planes, kernel_size, stride, padding, groups=groups, bias=False),
+            nn.BatchNorm2d(out_planes),
+            nn.ReLU6(inplace=True)
+        )
+
+
+class InvertedResidual(nn.Module):
+    def __init__(self, inp, oup, stride, expand_ratio):
+        super(InvertedResidual, self).__init__()
+        self.stride = stride
+        assert stride in [1, 2]
+
+        hidden_dim = int(round(inp * expand_ratio))
+        self.use_res_connect = self.stride == 1 and inp == oup
+
+        layers = []
+        if expand_ratio != 1:
+            # pw
+            layers.append(ConvBNReLU(inp, hidden_dim, kernel_size=1))
+        layers.extend([
+            # dw
+            ConvBNReLU(hidden_dim, hidden_dim, stride=stride, groups=hidden_dim),
+            # pw-linear
+            nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False),
+            nn.BatchNorm2d(oup),
+        ])
+        self.conv = nn.Sequential(*layers)
+
+    def forward(self, x):
+        if self.use_res_connect:
+            return x + self.conv(x)
+        else:
+            return self.conv(x)
+
+
+class MobileNetV2(nn.Module):
+    def __init__(self, num_classes=1000, width_mult=1.0, inverted_residual_setting=None, round_nearest=8):
+        """
+        MobileNet V2 main class
+        Args:
+            num_classes (int): Number of classes
+            width_mult (float): Width multiplier - adjusts number of channels in each layer by this amount
+            inverted_residual_setting: Network structure
+            round_nearest (int): Round the number of channels in each layer to be a multiple of this number
+            Set to 1 to turn off rounding
+        """
+        super(MobileNetV2, self).__init__()
+        block = InvertedResidual
+        input_channel = 64
+
+        if inverted_residual_setting is None:
+            inverted_residual_setting = [
+                # t, c, n, s               
+                [2, 32, 1, 1],
+                [2, 32, 1, 1],
+                [2, 64, 1, 2],
+                [2, 64, 1, 1],              
+            ]
+
+        # only check the first element, assuming user knows t,c,n,s are required
+        if len(inverted_residual_setting) == 0 or len(inverted_residual_setting[0]) != 4:
+            raise ValueError("inverted_residual_setting should be non-empty "
+                             "or a 4-element list, got {}".format(inverted_residual_setting))
+
+        # building first layer
+        input_channel = _make_divisible(input_channel * width_mult, round_nearest)
+        self.layers = []
+        # building inverted residual blocks
+        for t, c, n, s in inverted_residual_setting:
+            output_channel = _make_divisible(c * width_mult, round_nearest)
+            for i in range(n):
+                stride = s if i == 0 else 1
+                self.layers.append(block(input_channel, output_channel, stride, expand_ratio=t))
+                input_channel = output_channel
+
+        # weight initialization
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out')
+                if m.bias is not None:
+                    nn.init.zeros_(m.bias)
+            elif isinstance(m, nn.BatchNorm2d):
+                nn.init.ones_(m.weight)
+                nn.init.zeros_(m.bias)
+            elif isinstance(m, nn.Linear):
+                nn.init.normal_(m.weight, 0, 0.01)
+                nn.init.zeros_(m.bias)
+
+
+
+
+def multibox(mobilenet, num_classes):
+    loc_layers = []
+    conf_layers = []
+
+    loc_layers += [nn.Conv2d(32, 4, kernel_size=3, padding=1)]
+    conf_layers += [nn.Conv2d(32, 3 + (num_classes-1), kernel_size=3, padding=1)]
+
+    loc_layers += [nn.Conv2d(32, 4, kernel_size=3, padding=1)]
+    conf_layers += [nn.Conv2d(32, num_classes, kernel_size=3, padding=1)]
+
+    loc_layers += [nn.Conv2d(64, 4, kernel_size=3, padding=1)]
+    conf_layers += [nn.Conv2d(64, num_classes, kernel_size=3, padding=1)]
+
+    loc_layers += [nn.Conv2d(64, 4, kernel_size=3, padding=1)]
+    conf_layers += [nn.Conv2d(64, num_classes, kernel_size=3, padding=1)]
+
+      
+    return mobilenet, (loc_layers, conf_layers)
+
+
+def build_s3fd(phase, num_classes=2):
+    base_, head_ = multibox(
+        MobileNetV2().layers, num_classes)
+    
+    return S3FD(phase, base_, head_, num_classes)
+
+
+if __name__ == '__main__':
+    net = build_s3fd('train', num_classes=2)
+    inputs = Variable(torch.randn(4, 1, 320, 320))
+    output = net(inputs)

examples/pytorch/vision/Face_Detection/models/RPool_Face_QVGA_monochrome.py

@@ -172,7 +172,7 @@ class S3FD(nn.Module):
                 conf.view(conf.size(0), -1, self.num_classes),
                 self.priors
             )
-        return output
+        return output, loc, conf
 
     def load_weights(self, base_file):
         other, ext = os.path.splitext(base_file)
@@ -289,7 +289,6 @@ class MobileNetV2(nn.Module):
 
         # building first layer
         input_channel = _make_divisible(input_channel * width_mult, round_nearest)
-        self.last_channel = _make_divisible(last_channel * max(1.0, width_mult), round_nearest)
         self.layers = []
         # building inverted residual blocks
         for t, c, n, s in inverted_residual_setting:

examples/pytorch/vision/Face_Detection/models/RPool_Face_Quant.py

@@ -180,7 +180,7 @@ class S3FD(nn.Module):
                 conf.view(conf.size(0), -1, self.num_classes),
                 self.priors
             )
-        return output
+        return output, loc, conf
 
     def load_weights(self, base_file):
         other, ext = os.path.splitext(base_file)
@@ -297,7 +297,6 @@ class MobileNetV2(nn.Module):
 
         # building first layer
         input_channel = _make_divisible(input_channel * width_mult, round_nearest)
-        self.last_channel = _make_divisible(last_channel * max(1.0, width_mult), round_nearest)
         self.layers = []
         # building inverted residual blocks
         for t, c, n, s in inverted_residual_setting:

examples/pytorch/vision/Face_Detection/models/__init__.py

@@ -4,3 +4,5 @@
 from __future__ import absolute_import
 from .RPool_Face_C import *
 from .RPool_Face_Quant import *
+from .RPool_Face_QVGA_monochrome import *
+from .RPool_Face_M4 import *

examples/pytorch/vision/Face_Detection/prepare_wider_data.py

@@ -10,7 +10,10 @@ import os
 from data.config import cfg
 import cv2
 
-WIDER_ROOT = os.path.join(cfg.HOME, 'WIDER_FACE')
+HOME = os.environ['DATA_HOME']
+
+
+WIDER_ROOT = os.path.join(HOME, 'WIDER_FACE')
 train_list_file = os.path.join(WIDER_ROOT, 'wider_face_split',
                                'wider_face_train_bbx_gt.txt')
 val_list_file = os.path.join(WIDER_ROOT, 'wider_face_split',

examples/pytorch/vision/Face_Detection/scut_evaluation.py

@@ -0,0 +1,219 @@
+"""
+WiderFace evaluation code
+author: wondervictor
+mail: tianhengcheng@gmail.com
+copyright@wondervictor
+Code buit upon https://github.com/wondervictor/WiderFace-Evaluation having MIT License
+"""
+
+import os
+import tqdm
+import pickle
+import argparse
+import numpy as np
+from scipy.io import loadmat
+from bbox import bbox_overlaps
+from IPython import embed
+import cv2
+import sys
+
+sys.path.append('../')
+from data.choose_config import cfg
+cfg = cfg.cfg
+
+HOME = os.environ['DATA_HOME']
+SCUT_ROOT = os.path.join(HOME, 'SCUT_HEAD_Part_B')
+
+def read_pred_file(filepath):
+
+    with open(filepath, 'r') as f:
+        lines = f.readlines()
+        img_file = lines[0].rstrip('\n\r')
+        lines = lines[2:]
+
+    boxes = np.array(list(map(lambda x: [float(a) for a in x.rstrip('\r\n').split(' ')], lines))).astype('float')
+    return img_file.split('/')[-1], boxes
+
+
+def get_preds(pred_dir):
+    events = os.listdir(pred_dir)
+    boxes = dict()
+    pbar = tqdm.tqdm(events)
+
+    for event in pbar:
+        pbar.set_description('Reading Predictions ')
+        event_dir = os.path.join(pred_dir, event)
+        event_images = os.listdir(event_dir)
+        current_event = dict()
+        for imgtxt in event_images:
+            imgname, _boxes = read_pred_file(os.path.join(event_dir, imgtxt))
+            current_event[imgname.rstrip('.jpg')] = _boxes
+        boxes[event] = current_event
+    return boxes
+
+
+def norm_score(pred):
+    """ norm score
+    pred {key: [[x1,y1,x2,y2,s]]}
+    """
+
+    max_score = 0
+    min_score = 1
+
+    for _, k in pred.items():
+        for _, v in k.items():
+            if len(v) == 0:
+                continue
+            _min = np.min(v[:, -1])
+            _max = np.max(v[:, -1])
+            max_score = max(_max, max_score)
+            min_score = min(_min, min_score)
+
+    diff = max_score - min_score
+    for _, k in pred.items():
+        for _, v in k.items():
+            if len(v) == 0:
+                continue
+            v[:, -1] = (v[:, -1] - min_score)/diff
+
+
+def image_eval(pred, gt, iou_thresh):
+    """ single image evaluation
+    pred: Nx5
+    gt: Nx4
+    ignore:
+    """
+
+    _pred = pred.copy()
+    _gt = gt.copy()
+    pred_recall = np.zeros(_pred.shape[0])
+    recall_list = np.zeros(_gt.shape[0])
+    proposal_list = np.ones(_pred.shape[0])
+
+    _pred[:, 2] = _pred[:, 2] + _pred[:, 0]
+    _pred[:, 3] = _pred[:, 3] + _pred[:, 1]
+    _gt[:, 2] = _gt[:, 2] + _gt[:, 0]
+    _gt[:, 3] = _gt[:, 3] + _gt[:, 1]
+
+    overlaps = bbox_overlaps(_pred[:, :4], _gt)
+
+    for h in range(_pred.shape[0]):
+
+        gt_overlap = overlaps[h]
+        max_overlap, max_idx = gt_overlap.max(), gt_overlap.argmax()
+        if max_overlap >= iou_thresh:
+            if recall_list[max_idx] == 0:
+                recall_list[max_idx] = 1
+
+        r_keep_index = np.where(recall_list == 1)[0]
+        pred_recall[h] = len(r_keep_index)
+    return pred_recall, proposal_list
+
+
+def img_pr_info(thresh_num, pred_info, proposal_list, pred_recall):
+    pr_info = np.zeros((thresh_num, 2)).astype('float')
+    for t in range(thresh_num):
+
+        thresh = 1 - (t+1)/thresh_num
+        r_index = np.where(pred_info[:, 4] >= thresh)[0]
+        if len(r_index) == 0:
+            pr_info[t, 0] = 0
+            pr_info[t, 1] = 0
+        else:
+            r_index = r_index[-1]
+            p_index = np.where(proposal_list[:r_index+1] == 1)[0]
+            pr_info[t, 0] = len(p_index)
+            pr_info[t, 1] = pred_recall[r_index]
+    return pr_info
+
+
+def dataset_pr_info(thresh_num, pr_curve, count_face):
+    _pr_curve = np.zeros((thresh_num, 2))
+    for i in range(thresh_num):
+        _pr_curve[i, 0] = pr_curve[i, 1] / pr_curve[i, 0]
+        _pr_curve[i, 1] = pr_curve[i, 1] / count_face
+    return _pr_curve
+
+
+def voc_ap(rec, prec):
+
+    # correct AP calculation
+    # first append sentinel values at the end
+    mrec = np.concatenate(([0.], rec, [1.]))
+    mpre = np.concatenate(([0.], prec, [0.]))
+
+    # compute the precision envelope
+    for i in range(mpre.size - 1, 0, -1):
+        mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i])
+
+    # to calculate area under PR curve, look for points
+    # where X axis (recall) changes value
+    i = np.where(mrec[1:] != mrec[:-1])[0]
+
+    # and sum (\Delta recall) * prec
+    ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1])
+    return ap
+
+
+def evaluation(pred, iou_thresh=0.5):
+    f = open('../data/face_val_scutB.txt')
+    lines = f.readlines()
+    names = []
+    gt_list = []
+    for line in lines:
+        line = line.strip().split()
+        names.append(os.path.split(line[0])[1][0:-4])
+        th, tw, tc = cv2.imread(SCUT_ROOT + '/' + line[0]).shape
+        num_faces = int(line[1])
+        faces = []
+        for i in range(num_faces):
+            x = float(line[2 + 5 * i])
+            y = float(line[3 + 5 * i])
+            w = float(line[4 + 5 * i])
+            h = float(line[5 + 5 * i])
+
+            faces.append([x, y, w, h])
+        gt_list.append(faces)
+
+    thresh_num = 1000
+
+    aps = []
+
+    count_face = 0
+    pr_curve = np.zeros((thresh_num, 2)).astype('float')
+
+    for j in range(len(names)):
+        gt_boxes = np.asarray(gt_list[j]).astype('float')
+        pred_file = open(os.path.join(pred, names[j] + '.txt'))
+        pred_lines = pred_file.readlines()
+        pred_info = np.array(list(map(lambda x: [float(a) for a in x.rstrip('\r\n').split(' ')], pred_lines[2:]))).astype('float')
+    
+
+        count_face += len(gt_boxes)
+
+        if len(gt_boxes) == 0 or len(pred_info) == 0:
+            continue
+
+        pred_recall, proposal_list = image_eval(pred_info, gt_boxes, iou_thresh)
+
+        _img_pr_info = img_pr_info(thresh_num, pred_info, proposal_list, pred_recall)
+
+        pr_curve += _img_pr_info
+    pr_curve = dataset_pr_info(thresh_num, pr_curve, count_face)
+
+    propose = pr_curve[:, 0]
+    recall = pr_curve[:, 1]
+
+    ap = voc_ap(recall, propose)
+    aps.append(ap)
+    print(ap)
+
+
+
+if __name__ == '__main__':
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('-p', '--pred')
+
+    args = parser.parse_args()
+    evaluation(args.pred)

examples/pytorch/vision/Face_Detection/scut_test.py

@@ -0,0 +1,244 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT license.
+
+import os
+import argparse
+import torch
+import torch.nn as nn
+import torch.utils.data as data
+import torch.backends.cudnn as cudnn
+import torchvision.transforms as transforms
+import os.path as osp
+
+import cv2
+import time
+import numpy as np
+from PIL import Image
+import scipy.io as sio
+
+from data.choose_config import cfg
+cfg = cfg.cfg
+
+from torch.autograd import Variable
+from utils.augmentations import to_chw_bgr
+
+from importlib import import_module
+
+import warnings
+warnings.filterwarnings("ignore")
+
+HOME = os.environ['DATA_HOME']
+SCUT_ROOT = os.path.join(HOME, 'SCUT_HEAD_Part_B')
+
+def str2bool(v):
+    return v.lower() in ("yes", "true", "t", "1")
+
+parser = argparse.ArgumentParser(description='s3fd evaluatuon wider')
+parser.add_argument('--model', type=str,
+                    default='./weights/rpool_face_m4.pth', help='trained model')
+parser.add_argument('--thresh', default=0.05, type=float,
+                    help='Final confidence threshold')
+parser.add_argument('--multigpu',
+                    default=False, type=str2bool,
+                    help='Specify whether model was trained with multigpu')
+parser.add_argument('--model_arch',
+                    default='RPool_Face_M4', type=str,
+                    choices=['RPool_Face_M4', 'RPool_Face_QVGA_monochrome'],
+                    help='choose architecture among rpool variants')
+parser.add_argument('--save_folder', type=str,
+                    default='rpool_face_predictions', help='folder for saving predictions')
+
+
+args = parser.parse_args()
+
+
+use_cuda = torch.cuda.is_available()
+
+if use_cuda:
+    torch.set_default_tensor_type('torch.cuda.FloatTensor')
+else:
+    torch.set_default_tensor_type('torch.FloatTensor')
+
+
+def detect_face(net, img, shrink):
+    if shrink != 1:
+        img = cv2.resize(img, None, None, fx=shrink, fy=shrink,
+                         interpolation=cv2.INTER_LINEAR)
+
+    x = to_chw_bgr(img)
+    x = x.astype('float32')
+    x -= cfg.img_mean
+    x = x[[2, 1, 0], :, :]
+
+    if cfg.IS_MONOCHROME == True:
+        x = 0.299 * x[0] + 0.587 * x[1] + 0.114 * x[2]
+        x = torch.from_numpy(x).unsqueeze(0).unsqueeze(0)
+    else:
+        x = torch.from_numpy(x).unsqueeze(0)
+
+    if use_cuda:
+        x = x.cuda()
+
+    y = net(x)
+    detections = y.data
+    detections = detections.cpu().numpy()
+
+    det_conf = detections[0, 1, :, 0]
+    det_xmin = img.shape[1] * detections[0, 1, :, 1] / shrink
+    det_ymin = img.shape[0] * detections[0, 1, :, 2] / shrink
+    det_xmax = img.shape[1] * detections[0, 1, :, 3] / shrink
+    det_ymax = img.shape[0] * detections[0, 1, :, 4] / shrink
+    det = np.column_stack((det_xmin, det_ymin, det_xmax, det_ymax, det_conf))
+
+    keep_index = np.where(det[:, 4] >= args.thresh)[0]
+    det = det[keep_index, :]
+
+    return det
+
+
+def multi_scale_test(net, image, max_im_shrink):
+    # shrink detecting and shrink only detect big face
+    st = 0.5 if max_im_shrink >= 0.75 else 0.5 * max_im_shrink
+    det_s = detect_face(net, image, st)
+    index = np.where(np.maximum(
+        det_s[:, 2] - det_s[:, 0] + 1, det_s[:, 3] - det_s[:, 1] + 1) > 30)[0]
+    det_s = det_s[index, :]
+
+    # enlarge one times
+    bt = min(2, max_im_shrink) if max_im_shrink > 1 else (
+        st + max_im_shrink) / 2
+    det_b = detect_face(net, image, bt)
+
+    # enlarge small image x times for small face
+    if max_im_shrink > 2:
+        bt *= 2
+        while bt < max_im_shrink:
+            det_b = np.row_stack((det_b, detect_face(net, image, bt)))
+            bt *= 2
+        det_b = np.row_stack((det_b, detect_face(net, image, max_im_shrink)))
+
+    # enlarge only detect small face
+    if bt > 1:
+        index = np.where(np.minimum(
+            det_b[:, 2] - det_b[:, 0] + 1, det_b[:, 3] - det_b[:, 1] + 1) < 100)[0]
+        det_b = det_b[index, :]
+    else:
+        index = np.where(np.maximum(
+            det_b[:, 2] - det_b[:, 0] + 1, det_b[:, 3] - det_b[:, 1] + 1) > 30)[0]
+        det_b = det_b[index, :]
+
+    return det_s, det_b
+
+
+def flip_test(net, image, shrink):
+    image_f = cv2.flip(image, 1)
+    det_f = detect_face(net, image_f, shrink)
+
+    det_t = np.zeros(det_f.shape)
+    det_t[:, 0] = image.shape[1] - det_f[:, 2]
+    det_t[:, 1] = det_f[:, 1]
+    det_t[:, 2] = image.shape[1] - det_f[:, 0]
+    det_t[:, 3] = det_f[:, 3]
+    det_t[:, 4] = det_f[:, 4]
+    return det_t
+
+
+def bbox_vote(det):
+    order = det[:, 4].ravel().argsort()[::-1]
+    det = det[order, :]
+    while det.shape[0] > 0:
+        # IOU
+        area = (det[:, 2] - det[:, 0] + 1) * (det[:, 3] - det[:, 1] + 1)
+        xx1 = np.maximum(det[0, 0], det[:, 0])
+        yy1 = np.maximum(det[0, 1], det[:, 1])
+        xx2 = np.minimum(det[0, 2], det[:, 2])
+        yy2 = np.minimum(det[0, 3], det[:, 3])
+        w = np.maximum(0.0, xx2 - xx1 + 1)
+        h = np.maximum(0.0, yy2 - yy1 + 1)
+        inter = w * h
+        o = inter / (area[0] + area[:] - inter)
+
+        # get needed merge det and delete these det
+        merge_index = np.where(o >= 0.3)[0]
+        det_accu = det[merge_index, :]
+        det = np.delete(det, merge_index, 0)
+
+        if merge_index.shape[0] <= 1:
+            continue
+        det_accu[:, 0:4] = det_accu[:, 0:4] * np.tile(det_accu[:, -1:], (1, 4))
+        max_score = np.max(det_accu[:, 4])
+        det_accu_sum = np.zeros((1, 5))
+        det_accu_sum[:, 0:4] = np.sum(
+            det_accu[:, 0:4], axis=0) / np.sum(det_accu[:, -1:])
+        det_accu_sum[:, 4] = max_score
+        try:
+            dets = np.row_stack((dets, det_accu_sum))
+        except:
+            dets = det_accu_sum
+
+    dets = dets[0:750, :]
+    return dets
+
+
+if __name__ == '__main__':
+    cfg.USE_NMS = False
+    module = import_module('models.' + args.model_arch)
+    net = module.build_s3fd('test', cfg.NUM_CLASSES)
+    
+    if args.multigpu == True:
+        net = torch.nn.DataParallel(net)
+
+    checkpoint_dict = torch.load(args.model)
+    model_dict = net.state_dict()
+    model_dict.update(checkpoint_dict) 
+    net.load_state_dict(model_dict)
+
+    net.eval()
+
+    if use_cuda:
+        net.cuda()
+        cudnn.benckmark = True
+
+    counter = 0
+
+    f = open('./data/face_val_scutB.txt')
+    lines = f.readlines()
+
+    if os.path.exists('./{}'.format(args.save_folder)) is False:
+        os.mkdir('./{}'.format(args.save_folder))
+
+    for line in lines:
+        line = line.strip().split()
+        im_name = SCUT_ROOT + '/' + line[0]
+
+        img = Image.open(im_name)
+        img = img.convert('RGB')
+        img = np.array(img)
+
+        max_im_shrink = np.sqrt(
+            320 * 240 / (img.shape[0] * img.shape[1]))
+
+        shrink = max_im_shrink if max_im_shrink < 1 else 1
+        counter += 1
+
+        t1 = time.time()
+        det0 = detect_face(net, img, shrink)
+
+        dets = det0
+
+        t2 = time.time()
+        print('Detect %04d th image costs %.4f' % (counter, t2 - t1))
+
+        imgname = osp.split(line[0])[1]
+        save_path = './{}'.format(args.save_folder)
+        fout = open(osp.join(save_path, imgname[0:-4] + '.txt'), 'w')
+        fout.write('{:s}\n'.format(imgname[0:-4] + '.txt'))
+        fout.write('{:d}\n'.format(dets.shape[0]))
+        for i in range(dets.shape[0]):
+            xmin = dets[i][0]
+            ymin = dets[i][1]
+            xmax = dets[i][2]
+            ymax = dets[i][3]
+            score = dets[i][4]
+            fout.write('{:.1f} {:.1f} {:.1f} {:.1f} {:.3f}\n'.
+                       format(xmin, ymin, (xmax - xmin + 1), (ymax - ymin + 1), score))

examples/pytorch/vision/Face_Detection/train.py

@@ -29,9 +29,6 @@ parser.add_argument('--dataset',
                     default='face',
                     choices=['hand', 'face', 'head'],
                     help='Train target')
-parser.add_argument('--basenet',
-                    default='vgg16_reducedfc.pth',
-                    help='Pretrained base model')
 parser.add_argument('--batch_size',
                     default=16, type=int,
                     help='Batch size for training')
@@ -40,8 +37,11 @@ parser.add_argument('--resume',
                     help='Checkpoint state_dict file to resume training from')
 parser.add_argument('--model_arch',
                     default='RPool_Face_C', type=str,
-                    choices=['RPool_Face_C', 'RPool_Face_Quant', 'RPool_Face_QVGA_monochrome'],
+                    choices=['RPool_Face_C', 'RPool_Face_Quant', 'RPool_Face_QVGA_monochrome', 'RPool_Face_M4'],
                     help='choose architecture among rpool variants')
+parser.add_argument('--finetune',
+                    default=False, type=str2bool,
+                    help='Use CUDA to train model')
 parser.add_argument('--num_workers',
                     default=128, type=int,
                     help='Number of workers used in dataloading')
@@ -88,9 +88,13 @@ else:
 if not os.path.exists(args.save_folder):
     os.makedirs(args.save_folder)
 
+if args.finetune==True:
+    train_dataset = WIDERDetection('./data/face_train_scutB.txt', mode='train', mono_mode=cfg.IS_MONOCHROME, is_scut=True)
+    val_dataset = WIDERDetection('./data/face_val_scutB.txt', mode='val', mono_mode=cfg.IS_MONOCHROME, is_scut=True)
+else:
+    train_dataset = WIDERDetection(cfg.FACE.TRAIN_FILE, mode='train', mono_mode=cfg.IS_MONOCHROME)
+    val_dataset = WIDERDetection(cfg.FACE.VAL_FILE, mode='val', mono_mode=cfg.IS_MONOCHROME)
 
-train_dataset = WIDERDetection(cfg.FACE.TRAIN_FILE, mode='train', mono_mode=cfg.IS_MONOCHROME)
-val_dataset = WIDERDetection(cfg.FACE.VAL_FILE, mode='val', mono_mode=cfg.IS_MONOCHROME)
 
 train_loader = data.DataLoader(train_dataset, args.batch_size,
                                num_workers=args.num_workers,
@@ -112,7 +116,6 @@ module = import_module('models.' + args.model_arch)
 net = module.build_s3fd('train', cfg.NUM_CLASSES)
 
 
-
 if args.cuda:
     if args.multigpu:
         net = torch.nn.DataParallel(net)
@@ -127,7 +130,7 @@ optimizer = optim.SGD(net.parameters(), lr=args.lr, momentum=args.momentum,
                       weight_decay=args.weight_decay)
 
 criterion = MultiBoxLoss(cfg, args.dataset, args.cuda)
-print('Loading wider dataset...')
+print('Loading dataset...')
 print('Using the specified args:')
 print(args)
 
@@ -153,7 +156,7 @@ def train():
 
               
             t0 = time.time()
-            out = net(images)
+            out,_,_ = net(images)
             # backprop
             optimizer.zero_grad()
             loss_l, loss_c = criterion(out, targets)
@@ -179,9 +182,16 @@ def train():
                            os.path.join(args.save_folder, file))
             iteration += 1
 
+            if args.model_arch == 'RPool_Face_M4' and args.finetune == True:
+                if args.multigpu == True:
+                    net.module.rnn_model.cell_rnn.cell.sparsify()
+                    net.module.rnn_model.cell_bidirrnn.cell.sparsify()
+                else:
+                    net.rnn_model.cell_rnn.cell.sparsify()
+                    net.rnn_model.cell_bidirrnn.cell.sparsify()
+                net.to('cuda')
+
         val(epoch)
-        if iteration == cfg.MAX_STEPS:
-            break
 
 
 def val(epoch):
@@ -200,7 +210,7 @@ def val(epoch):
                 images = images
                 targets = [ann for ann in targets]
 
-            out = net(images)
+            out,_,_ = net(images)
             loss_l, loss_c = criterion(out, targets)
             loss = loss_l + loss_c
             loc_loss += loss_l.item()
@@ -241,4 +251,4 @@ def adjust_learning_rate(optimizer, epoch, iteration, num_iter):
 
 
 if __name__ == '__main__':
-    train()
+    train()

examples/pytorch/vision/Face_Detection/utils/augmentations.py

@@ -775,6 +775,87 @@ def anchor_crop_image_sampling(img,
 
         return image, sampled_labels
 
+def reduce_image(img, bbox_labels, img_width, img_height):
+    bbox_labels = np.array(bbox_labels)
+    scale = np.array([img_width, img_height, img_width, img_height])
+
+    boxes = bbox_labels[:, 1:5] * scale
+
+    boxArea = (boxes[:, 2] - boxes[:, 0] + 1) * (boxes[:, 3] - boxes[:, 1] + 1)
+    boxArea_max = np.amax(boxArea)
+
+    if boxArea_max > 48*48:
+        reduce_ratio = (boxArea_max/(48*48))**(0.5)
+        height = int(img_height / reduce_ratio)
+        width = int(img_width / reduce_ratio)
+
+        img = img.resize((width, height),
+                     resample=Image.LANCZOS)
+
+        boxes = boxes//reduce_ratio
+        new_scale = np.array([width, height, width, height])
+        boxes = (boxes[:, 0:4]/new_scale)
+
+        bbox_labels[:,1:5] = boxes
+
+        bbox_labels = bbox_labels.tolist()
+
+        img_height = height
+        img_width = width
+
+        ratio_h = img_height/320
+        ratio_w = img_width/320
+        if ratio_h > 1 and ratio_w > 1:
+            expand_ratio = 1
+        else: 
+            expand_ratio = 1/(min(ratio_h,ratio_w))
+
+        height = int(img_height * expand_ratio)
+        width = int(img_width * expand_ratio)
+        h_off = math.floor(np.random.uniform(0, height - img_height))
+        w_off = math.floor(np.random.uniform(0, width - img_width))
+        expand_bbox = bbox(-w_off / img_width, -h_off / img_height,
+                           (width - w_off) / img_width,
+                           (height - h_off) / img_height)
+        expand_img = np.ones((height, width, 3))
+        expand_img = np.uint8(expand_img * np.squeeze(cfg.img_mean))
+        expand_img = Image.fromarray(expand_img)
+        expand_img.paste(img, (int(w_off), int(h_off)))
+        bbox_labels = transform_labels(bbox_labels, expand_bbox)
+        return expand_img, bbox_labels, width, height
+    return img, bbox_labels.tolist(), img_width, img_height
+
+
+def precrop(img, bbox_labels):
+    img_height, img_width = img.size[1], img.size[0]
+    scale = np.array([img_width, img_height, img_width, img_height])
+    bbox_labels = np.array(bbox_labels)
+    bbox_labels[:, 1:5] = bbox_labels[:, 1:5] * scale
+
+    if img_width-20 > img_height:
+        height_max, height_min = img_width + 20, img_width - 20
+        new_height = np.random.randint(height_min, height_max)
+        expand_img = np.ones((new_height, img_width, 3))
+        expand_img = np.uint8(expand_img * np.squeeze(cfg.img_mean))
+        expand_img = Image.fromarray(expand_img)
+        expand_img.paste(img, (int(0), int(0)))
+        img_height, img_width = new_height, img_width
+        img = expand_img
+
+    elif img_height-20 > img_width:
+        width_max, width_min = img_height + 20, img_height - 20
+        new_width = np.random.randint(width_min, width_max)
+        expand_img = np.ones((img_height, new_width, 3))
+        expand_img = np.uint8(expand_img * np.squeeze(cfg.img_mean))
+        expand_img = Image.fromarray(expand_img)
+        expand_img.paste(img, (int(0), int(0)))
+        img_height, img_width = img_height, new_width
+        img = expand_img
+
+    scale = np.array([img_width, img_height, img_width, img_height])
+    bbox_labels[:, 1:5] = bbox_labels[:, 1:5] / scale
+    return img, bbox_labels.tolist(), img_width, img_height
+
 
 def preprocess(img, bbox_labels, mode, image_path):
     img_width, img_height = img.size
@@ -860,3 +941,51 @@ def preprocess(img, bbox_labels, mode, image_path):
     #img = img * cfg.scale
 
     return img, sampled_labels
+
+def preprocess_qvga(img, bbox_labels, mode, image_path):
+    img_width, img_height = img.size
+    sampled_labels = bbox_labels
+    if mode == 'train':
+
+        img, bbox_labels, img_width, img_height = precrop(img, bbox_labels)
+
+        if cfg.apply_distort:
+            img = distort_image(img)
+
+        scale_array = np.array([8, 16, 32, 48])
+
+        img, bbox_labels, img_width, img_height = reduce_image(
+                img, bbox_labels, img_width, img_height)
+
+        img = np.array(img)
+        sampled_labels = bbox_labels
+
+        img = img.astype('uint8')
+        img = Image.fromarray(img)
+
+    interp_mode = [
+        Image.BILINEAR, Image.HAMMING, Image.NEAREST, Image.BICUBIC,
+        Image.LANCZOS
+    ]
+    interp_indx = np.random.randint(0, 5)
+
+    img = img.resize((cfg.resize_width, cfg.resize_height),
+                     resample=interp_mode[interp_indx])
+
+    img = np.array(img)
+
+    if mode == 'train':
+        mirror = int(np.random.uniform(0, 2))
+        if mirror == 1:
+            img = img[:, ::-1, :]
+            for i in six.moves.xrange(len(sampled_labels)):
+                tmp = sampled_labels[i][1]
+                sampled_labels[i][1] = 1 - sampled_labels[i][3]
+                sampled_labels[i][3] = 1 - tmp
+
+    img = to_chw_bgr(img)
+    img = img.astype('float32')
+    img -= cfg.img_mean
+    img = img[[2, 1, 0], :, :]  # to RGB
+
+    return img, sampled_labels

examples/pytorch/vision/Face_Detection/wider_test.py

@@ -27,11 +27,19 @@ from importlib import import_module
 import warnings
 warnings.filterwarnings("ignore")
 
+HOME = os.environ['DATA_HOME']
+
+def str2bool(v):
+    return v.lower() in ("yes", "true", "t", "1")
+
 parser = argparse.ArgumentParser(description='s3fd evaluatuon wider')
 parser.add_argument('--model', type=str,
                     default='./weights/rpool_face_c.pth', help='trained model')
 parser.add_argument('--thresh', default=0.05, type=float,
                     help='Final confidence threshold')
+parser.add_argument('--multigpu',
+                    default=False, type=str2bool,
+                    help='Specify whether model was trained with multigpu')
 parser.add_argument('--model_arch',
                     default='RPool_Face_C', type=str,
                     choices=['RPool_Face_C', 'RPool_Face_Quant', 'RPool_Face_QVGA_monochrome'],
@@ -73,7 +81,7 @@ def detect_face(net, img, shrink):
     if use_cuda:
         x = x.cuda()
 
-    y = net(x)
+    y,_,_ = net(x)
     detections = y.data
     detections = detections.cpu().numpy()
 
@@ -177,7 +185,7 @@ def bbox_vote(det):
 def get_data():
     subset = args.subset
 
-    WIDER_ROOT = os.path.join(cfg.HOME, 'WIDER_FACE')
+    WIDER_ROOT = os.path.join(HOME, 'WIDER_FACE')
     if subset == 'val':
         wider_face = sio.loadmat(
             os.path.join(WIDER_ROOT, 'wider_face_split',
@@ -191,7 +199,7 @@ def get_data():
     del wider_face
 
     imgs_path = os.path.join(
-        cfg.FACE.WIDER_DIR, 'WIDER_{}'.format(subset), 'images')
+        WIDER_ROOT, 'WIDER_{}'.format(subset), 'images')
     save_path = './{}'.format(args.save_folder)
 
     return event_list, file_list, imgs_path, save_path
@@ -203,7 +211,8 @@ if __name__ == '__main__':
     module = import_module('models.' + args.model_arch)
     net = module.build_s3fd('test', cfg.NUM_CLASSES)
     
-    net = torch.nn.DataParallel(net)
+    if args.multigpu == True:
+        net = torch.nn.DataParallel(net)
     
 
     checkpoint_dict = torch.load(args.model)

pytorch/edgeml_pytorch/graph/rnn.py

@@ -186,7 +186,9 @@ class RNNCell(nn.Module):
             mats[i] = utils.hardThreshold(mats[i], self._wSparsity)
         for i in range(endW, endU):
             mats[i] = utils.hardThreshold(mats[i], self._uSparsity)
-        self.copy_previous_UW()
+        self.W.data.copy_(mats[0])
+        self.U.data.copy_(mats[1])
+        # self.copy_previous_UW()
 
     def sparsifyWithSupport(self):
         mats = self.getVars()

pytorch/edgeml_pytorch/graph/rnnpool.py

@@ -8,13 +8,18 @@ from edgeml_pytorch.graph.rnn import *
 
 class RNNPool(nn.Module):
     def __init__(self, nRows, nCols, nHiddenDims,
-                     nHiddenDimsBiDir, inputDims):
+                     nHiddenDimsBiDir, inputDims, 
+                     w1Sparsity=1.0, u1Sparsity=1.0, w2Sparsity=1.0, u2Sparsity=1.0):
         super(RNNPool, self).__init__()
         self.nRows = nRows
         self.nCols = nCols
         self.inputDims = inputDims
         self.nHiddenDims = nHiddenDims
         self.nHiddenDimsBiDir = nHiddenDimsBiDir
+        self.w1Sparsity = w1Sparsity
+        self.u1Sparsity = u1Sparsity
+        self.w2Sparsity = w2Sparsity
+        self.u2Sparsity = u2Sparsity
 
         self._build()
 
@@ -22,11 +27,13 @@ class RNNPool(nn.Module):
 
         self.cell_rnn = FastGRNN(self.inputDims, self.nHiddenDims, gate_nonlinearity="sigmoid",
                                 update_nonlinearity="tanh", zetaInit=100.0, nuInit=-100.0,
-                                batch_first=False, bidirectional=False)
+                                batch_first=False, bidirectional=False, 
+                                wSparsity=self.w1Sparsity, uSparsity=self.u1Sparsity)
 
         self.cell_bidirrnn = FastGRNN(self.nHiddenDims, self.nHiddenDimsBiDir, gate_nonlinearity="sigmoid",
                                 update_nonlinearity="tanh", zetaInit=100.0, nuInit=-100.0,
-                                batch_first=False, bidirectional=True, is_shared_bidirectional=True)
+                                batch_first=False, bidirectional=True, is_shared_bidirectional=True,
+                                wSparsity=self.w2Sparsity, uSparsity=self.u2Sparsity)
 
 
     def static_single(self,inputs, hidden, batch_size):
@@ -40,24 +47,24 @@ class RNNPool(nn.Module):
         row_timestack = torch.cat(torch.unbind(inputs, dim=3),dim=0) 
 
         stateList = self.static_single(torch.stack(torch.unbind(row_timestack,dim=2)),
-                        (torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(torch.device("cuda")),
-                        torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(torch.device("cuda"))),batch_size)       
+                        (torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(inputs.device),
+                        torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(inputs.device)),batch_size)       
 
         outputs_cols = self.cell_bidirrnn(torch.stack(stateList),
-                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(torch.device("cuda")),
-                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(torch.device("cuda")))
+                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(inputs.device),
+                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(inputs.device))
 
 
         ## across columns
         col_timestack = torch.cat(torch.unbind(inputs, dim=2),dim=0)
 
         stateList = self.static_single(torch.stack(torch.unbind(col_timestack,dim=2)),
-                        (torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(torch.device("cuda")),
-                        torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(torch.device("cuda"))),batch_size)
+                        (torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(inputs.device),
+                        torch.zeros(1, batch_size * self.nRows, self.nHiddenDims).to(inputs.device)),batch_size)
 
         outputs_rows = self.cell_bidirrnn(torch.stack(stateList),
-                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(torch.device("cuda")),
-                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(torch.device("cuda")))
+                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(inputs.device),
+                        torch.zeros(2, batch_size, self.nHiddenDimsBiDir).to(inputs.device))