Home

Awesome

DecoupleGCN-DropGraph

The implementation for "Decoupling GCN with DropGraph Module for Skeleton-Based Action Recognition" (ECCV2020). The proposed method boosts the performance of spatial-temporal graph convolutional network with NO extra FLOPs, NO extra latency, and NO extra GPU memory cost.

Prerequisite

Data Preparation

Training & Testing

Multi-stream ensemble

To ensemble the results of 4 streams. Change models name in ensemble.py depending on your experiment setting. Then run python ensemble.py.

Trained models

We release several trained models:

ModelDatasetSettingTop1(%)
./save_models/ntu_joint_xview.ptNTU-RGBDX-view95.2
./save_models/ntu_joint_xsub.ptNTU-RGBDX-sub88.2
./save_models/ntu120_joint_xsetup.ptNTU-RGBD120X-setup84.3
./save_models/ntu120_joint_xsub.ptNTU-RGBD120X-sub82.4

Citation

If you find this model useful for your resesarch, please use the following BibTeX entry.

@inproceedings{cheng2020eccv,  
  title     = {Decoupling GCN with DropGraph Module for Skeleton-Based Action Recognition},  
  author    = {Ke Cheng and Yifan Zhang and Congqi Cao and Lei Shi and Jian Cheng and Hanqing Lu},  
  booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},  
  year      = {2020},  
}