Update docs for unsupervised learning

README.md

@@ -21,26 +21,36 @@ By [Peng-Shuai Wang](https://wang-ps.github.io/), [Yang Liu](https://xueyuhanlan
 and [Xin Tong](https://www.microsoft.com/en-us/research/people/xtong/)<br/>
 Computer Vision and Pattern Recognition (CVPR) Workshops, 2020<br/>
 
+- **[Unsupervised 3D Learning for Shape Analysis via Multiresolution Instance Discrimination](https://arxiv.org/abs/2008.01068)**<br/>
+By [Peng-Shuai Wang](https://wang-ps.github.io/), Yu-Qi Yang, Qian-Fang Zou, 
+[Zhirong Wu](https://www.microsoft.com/en-us/research/people/wuzhiron/), 
+[Yang Liu](https://xueyuhanlang.github.io/) 
+and [Xin Tong](https://www.microsoft.com/en-us/research/people/xtong/)<br/>
+Arxiv preprint, 2020<br/>
+
+If you use our code or models, please [cite](docs/citation.md) our paper.
+
+
+### What's New?
+
+- 2020.08.16: We released our code for [3D unsupervised learning](docs/unsupervised.md).
+  We provided a unified network architecture for generic shape analysis tasks and 
+  an unsupervised method to pretrain the network. Our method achieved state-of-the-art 
+  performance on several benchmarks.
+- 2020.08.12: We released our code for 
+  [Partnet segmentation](docs/segmentation.md#shape-segmentation-on-partnet-with-tensorflow).
+  We achieved  an average IoU of **58.4**, significantly better than PointNet
+  (IoU: 35.6), PointNet++ (IoU: 42.5), SpiderCNN (IoU: 37.0), and PointCNN(IoU:
+  46.5).
+- 2020.08.05: We released our code for [shape completion](docs/completion.md).
+  We proposed a simple yet efficient network and output-guided skip connections
+  for 3D completion, which achieved state-of-the-art performances on several 
+  benchmarks.
+- 2020.03.16: We released ResNet-based O-CNN architecture for 
+  [shape classification](docs/classification.md#o-cnn-on-tensorflow).
+  We achieved a testing accuracy of **92.5** on ModelNet40 (without voting).
+
 
-If you use our code or models, please cite our paper.
-```
-@article {Wang-2017-OCNN,
-    title     = {{O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis}},
-    author    = {Wang, Peng-Shuai and Liu, Yang and Guo, Yu-Xiao and Sun, Chun-Yu and Tong, Xin},
-    journal   = {ACM Transactions on Graphics (SIGGRAPH)},
-    volume    = {36},
-    number    = {4},
-    year      = {2017},
-}
-@article {Wang-2018-AOCNN,
-    title     = {{Adaptive O-CNN: A Patch-based Deep Representation of 3D Shapes}},
-    author    = {Wang, Peng-Shuai and Sun, Chun-Yu and Liu, Yang and Tong, Xin},
-    journal   = {ACM Transactions on Graphics (SIGGRAPH Asia)},
-    volume    = {37},
-    number    = {6},
-    year      = {2018},
-}
-```
 
 ### Contents
 - [Installation](docs/installation.md)

docs/citation.md

@@ -0,0 +1,37 @@
+# Citations
+
+```
+@article {Wang-2017-OCNN,
+    title     = {{O-CNN}: Octree-based Convolutional Neural Networks for {3D} Shape Analysis},
+    author    = {Wang, Peng-Shuai and Liu, Yang and Guo, Yu-Xiao and Sun, Chun-Yu and Tong, Xin},
+    journal   = {ACM Transactions on Graphics (SIGGRAPH)},
+    volume    = {36},
+    number    = {4},
+    year      = {2017},
+}
+
+@article {Wang-2018-AOCNN,
+    title     = {{Adaptive O-CNN}: A Patch-based Deep Representation of {3D} Shapes},
+    author    = {Wang, Peng-Shuai and Sun, Chun-Yu and Liu, Yang and Tong, Xin},
+    journal   = {ACM Transactions on Graphics (SIGGRAPH Asia)},
+    volume    = {37},
+    number    = {6},
+    year      = {2018},
+}
+
+@InProceedings {Wang-2020-Completion,
+    title     = {Deep Octree-based {CNNs} with Output-Guided Skip Connections for {3D} Shape and Scene Completion},
+    author    = {Wang, Peng-Shuai and Liu, Yang and Tong, Xin},
+    journal   = {Computer Vision and Pattern Recognition (CVPR) Workshops},
+    year      = {2020},
+}
+
+@article{Wang-2020-Unsupervised,
+  title      = {Unsupervised {3D} Learning for Shape Analysis via Multiresolution Instance Discrimination},
+  author     = {Wang, Peng-Shuai and Yang, Yu-Qi and Zou, Qian-Fang and Wu, Zhirong and Liu, Yang and Tong, Xin},
+  journal    = {arXiv preprint arXiv:2008.01068},
+  year       = {2020},
+}
+
+
+```

docs/unsupervised.md

@@ -0,0 +1,32 @@
+# 3D Unsupervised Learning
+
+## Unsupervised pretraining
+
+1. Download the data and pretrained weights with the following command. 
+   ```shell
+   cd  tensorflow
+   python data/midnet_data.py
+   ```
+
+2. Run the following command to train the network. 
+   ```shell
+   cd  script
+   python run_mid.py --config configs/mid_hrnet_d6.yaml
+   ```
+   
+
+## Finetune on PartNet
+
+Follow the instructions 
+[here](docs/segmentation.md#shape-segmentation-on-partnet-with-tensorflow)
+to download the PartNet.
+
+Run the following script to finetune the network on PartNet with the pretrained
+weights we provided. Compared with a random initialization, the IoU increases from
+58.4 to 60.8. If you would like to finetune the network with your own pretrained
+weights, you can simply provide the checkpoint via the command parameter `--ckpt`.
+
+```shell
+cd script
+python run_seg_partnet_cmd.py --alias partnet_finetune --finetune
+```