- combine classification + filter visualization
- order by classification, learning LeNet, brewing logreg, and
fine-tuning to flickr style
- improve flow of content in classification + filter visualization
- include solver needed for learning LeNet
- edit notebook descriptions for site catalogue
This is a quick translation of the examples to the caffe.io.Transformer
interface. The results are not strictly identical to the earlier
implementation! The models now use a mean pixel instead of a mean image
for simplicity. The output classifications and detections are preserved
but scores may differ.
Note: the examples will be rewritten to make use of `caffe.Net` alone
since it is the true interface, but not yet.
- load an image as [0,1] single / np.float32 according to Python convention
- fix input scaling during preprocessing:
- scale input for preprocessing by `raw_scale` e.g. to map an image
to [0, 255] for the CaffeNet and AlexNet ImageNet models
- scale feature space by `input_scale` after mean subtraction
- switch examples to raw scale for ImageNet models
- fix#525
- preserve type after resizing.
- resize 1, 3, or K channel images with special casing between
skimage.transform (1 and 3) and scipy.ndimage (K) for speed
- run through and save new output
- collect region proposals with R-CNN configuration (see sergeyk/selective_search_ijcv_with_python)
- call detect.py in GPU mode
- fix NMS plotting: X and Y coords were accidentally exchanged. print scores too.
Evan Shelhamer
Sebastián Ramírez
PETER_GAO
Sergey Karayev