Awesome

For SwapNet

Create a list.txt file containing all the images to process. This can be done with the GNU find command:

find path/to/input/folder -name '*.jpg' -o -name '*.png' > list.txt

Then run this to get the clothing segmentations

python evaluate_parsing_JPPNet-s2.py -d path/to/texture -l path/to/list.txt -o path/to/clothing

Joint Body Parsing & Pose Estimation Network (JPPNet)

Xiaodan Liang, Ke Gong, Xiaohui Shen, and Liang Lin, "Look into Person: Joint Body Parsing & Pose Estimation Network and A New Benchmark", T-PAMI 2018.

Introduction

JPPNet is a state-of-art deep learning methord for human parsing and pose estimation built on top of Tensorflow.

This novel joint human parsing and pose estimation network incorporates the multiscale feature connections and iterative location refinement in an end-to-end framework to investigate efficient context modeling and then enable parsing and pose tasks that are mutually beneficial to each other. This unified framework achieves state-of-the-art performance for both human parsing and pose estimation tasks.

This distribution provides a publicly available implementation for the key model ingredients reported in our latest paper which is accepted by T-PAMI 2018.

We simplify the network to solve human parsing by exploring a novel self-supervised structure-sensitive learning approach, which imposes human pose structures into the parsing results without resorting to extra supervision. There is also a public implementation of this self-supervised structure-sensitive JPPNet (SS-JPPNet).

Look into People (LIP) Dataset

The SSL is trained and evaluated on our LIP dataset for human parsing. Please check it for more model details. The dataset is also available at google drive and baidu drive.

Pre-trained models

We have released our trained models of JPPNet on LIP dataset at google drive and baidu drive.

Inference

1. Download the pre-trained model and store in $HOME/checkpoint.
2. Prepare the images and store in $HOME/datasets.
3. Run evaluate_pose_JPPNet-s2.py for pose estimation and evaluate_parsing_JPPNet-s2.py for human parsing.
4. The results are saved in $HOME/output

Training

1. Download the pre-trained model and store in $HOME/checkpoint.
2. Download LIP dataset or prepare your own data and store in $HOME/datasets.
3. For LIP dataset, we have provided images, parsing labels, lists and the left-right flipping labels (labels_rev) for data augmentation. You need to generate the heatmaps of pose labels. We have provided a script for reference.
4. Run train_JPPNet-s2.py to train the JPPNet with two refinement stages.
5. Use evaluate_pose_JPPNet-s2.py and evaluate_parsing_JPPNet-s2.py to generate the results or evaluate the trained models.
6. Note that the LIPReader class is only suit for labels in LIP for the left-right flipping augmentation. If you want to train on other datasets with different labels, you may have to re-write an image reader class.

Citation

If you use this code for your research, please cite our papers.

@article{liang2018look,
  title={Look into Person: Joint Body Parsing \& Pose Estimation Network and a New Benchmark},
  author={Liang, Xiaodan and Gong, Ke and Shen, Xiaohui and Lin, Liang},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2018},
  publisher={IEEE}
}

@InProceedings{Gong_2017_CVPR,
  author = {Gong, Ke and Liang, Xiaodan and Zhang, Dongyu and Shen, Xiaohui and Lin, Liang},
  title = {Look Into Person: Self-Supervised Structure-Sensitive Learning and a New Benchmark for Human Parsing},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {July},
  year = {2017}
}