Simba

This is the official implementation of our paper Simba: Mamba augmented U-ShiftGCN for Skeletal Action Recognition in Videos. To the best of our knowledge, this is the first skeleton action recognition model based on Mamba augmentation with GCN.

News

• Make pre-trained models available.
• This repo borrows inspiration from multiple codebases. Therefore, some files are redundant. Working towards cleaning the repo !
• Make the code available.

Simba architecture

Preparation

Install torchlight

Run pip install -e torchlight

Install dependencies

pip install -r requirements.txt

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 CTR-GCN
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

Evaluation

bash evaluate.sh

Replace path_to_model_weights argument to the saved model weights you obtain after training.

Training & Testing

Training

bash train.sh

Please check the configuration in the config directory.

Testing

bash evaluate.sh

To ensemble the results of different modalities, run the following command:

bash ensemble.sh

Acknowledgements

This repo is based on Hyperformer, Mamba and Shift-GCN.

Thanks to the original authors for their work!

Citation

Please cite this work if you find it useful:

@article{chaudhuri2024simba,
  title={Simba: Mamba augmented U-ShiftGCN for Skeletal Action Recognition in Videos},
  author={Chaudhuri, Soumyabrata and Bhattacharya, Saumik},
  journal={arXiv preprint arXiv:2404.07645},
  year={2024}
}

Contact

For any questions, feel free to raise an issue !