FR-Head

This repo is the official implementation for Learning Discriminative Representations for Skeleton Based Action Recognition.

The paper is accepted to CVPR2023.

Architecture

Prerequisites

• Python >= 3.6

• PyTorch >= 1.1.0

• PyYAML, tqdm, tensorboardX

• We provide the dependency file of our experimental environment, you can install all dependencies by creating a new anaconda virtual environment and running pip install -r requirements.txt

• Run pip install -e torchlight

Data Preparation

Download datasets.

There are 3 datasets to download:

• NTU RGB+D 60 Skeleton
• NTU RGB+D 120 Skeleton
• NW-UCLA

NTU RGB+D 60 and 120

1. Request dataset here: https://rose1.ntu.edu.sg/dataset/actionRecognition
2. Download the skeleton-only datasets:
   1. nturgbd_skeletons_s001_to_s017.zip (NTU RGB+D 60)
   2. nturgbd_skeletons_s018_to_s032.zip (NTU RGB+D 120)
   3. Extract above files to ./data/nturgbd_raw

NW-UCLA

1. Download dataset from here
2. Move all_sqe to ./data/NW-UCLA

Data Processing

Directory Structure

Put downloaded data into the following directory structure:

- data/
  - NW-UCLA/
    - all_sqe
      ... # raw data of NW-UCLA
  - ntu/
  - ntu120/
  - nturgbd_raw/
    - nturgb+d_skeletons/     # from `nturgbd_skeletons_s001_to_s017.zip`
      ...
    - nturgb+d_skeletons120/  # from `nturgbd_skeletons_s018_to_s032.zip`
      ...

Generating Data

• Generate NTU RGB+D 60 or NTU RGB+D 120 dataset:

 cd ./data/ntu # or cd ./data/ntu120
 # Get skeleton of each performer
 python get_raw_skes_data.py
 # Remove the bad skeleton 
 python get_raw_denoised_data.py
 # Transform the skeleton to the center of the first frame
 python seq_transformation.py

Training & Testing

Pretrained Models & Results

• We provide pretrained models under ./results folder for producing the final results on all 5 benchmarks.
• Test the pretrained models directly or refer to config.yaml under the related subfolder to reproduce the model.
• The detailed results for provided models are displayed below:

Training

• Change the config file or assign the arguments using the command line depending on what you want.
• Refer to ./utils/config.py for detailed description of all arguments.

# Example: training the default backbone (CTR-GCN) without FR-Head on NTU RGB+D 60 cross subject with GPU 0
python main.py --config config/nturgbd-cross-subject/rand_scale.yaml --work-dir results/ntu/xsub/demo --device 0
# Example: training the default backbone (CTR-GCN) with FR-Head
python main.py --config config/nturgbd-cross-subject/rand_scale.yaml --cl-mode ST-Multi-Level --w-multi-cl-loss 0.1 0.2 0.5 1 --work-dir results/ntu/xsub/demo --device 0
# Example: training FR-Head based on other backbones
python main.py --config config/nturgbd-cross-subject/rand_scale.yaml --cl-mode ST-Multi-Level --w-multi-cl-loss 0.1 0.2 0.5 1 --work-dir results/ntu/xsub/demo --device 0

• To train model with other modalities, setting bone or vel arguments in the config file or use the existed config file.

# Example: training models on NTU RGB+D 120 cross subject under bone modality
python main.py --config config/nturgbd120-cross-subject/rand_scale_bone.yaml --work-dir results/ntu120/xsub/demo --cl-mode ST-Multi-Level --w-multi-cl-loss 0.1 0.2 0.5 1 --device 0

Testing

• To test the trained models saved in ./results, run the following command:

python main.py --config config/ucla/default.yaml --work-dir results/NW-UCLA/demo --phase test --save-score True --weights results/NW-UCLA/FR-Head_Joint_94.40/runs.pt --device 0

• To ensemble the results of different modalities, run

  # Example: ensemble four modalities of pretrained modelson NTU RGB+D 120 cross set
  python ensemble.py --dataset ntu120/xset --joint-dir results/ntu120/xset/FR-Head_Joint_87.32/ --bone-dir results/ntu120/xset/FR-Head_Bone_88.14/ --joint-motion-dir results/ntu120/xset/FR-Head_Joint_Motion_83.98/ --bone-motion-dir results/ntu120/xset/FR-Head_Bone_Motion_83.91/

Acknowledgements

This repo is mainly based on CTR-GCN. We also refer to ST-GCN, 2s-AGCN and TCA-GCN.

Thanks to the original authors for their excellent work!

Citation

Please cite this work if you find it useful:.

  @inproceedings{zhou2023learning,
     title={Learning Discriminative Representations for Skeleton Based Action Recognition},
     author={Zhou, Huanyu and Liu, Qingjie and Wang, Yunhong},
     booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
     year={2023}
  }