DB-LSTM: Densely-Connected Bi-directional LSTM for Human Action Recognition

partial code for paper "DB-LSTM: Densely-Connected Bi-directional LSTM for Human Action Recognition"

Requirements

• python3
• pytorch

Installation

mkdir $DYNETDIR
git clone https://github.com/zhiqicheng/DB-LSTM

Example command

Training:

python main.py

Testing:

python eval_score.py

Major changes:

• major refactoring of internal data handling

Citation

If you find our code useful, please consider citing:

@article{he2021db,
  title={DB-LSTM: Densely-connected Bi-directional LSTM for human action recognition},
  author={He, Jun-Yan and Wu, Xiao and Cheng, Zhi-Qi and Yuan, Zhaoquan and Jiang, Yu-Gang},
  journal={Neurocomputing},
  volume={444},
  pages={319--331},
  year={2021},
  publisher={Elsevier}
}