thraster / SGCN4HOI

SGCN4HOI: A Skeleton-aware Graph Convolutional Network for Human-Object Interaction Detection

SGCN4HOI

SGCN4HOI: A Skeleton-aware Graph Convolutional Network for Human-Object Interaction Detection (Paper)

Dataset & Requirements

Please follow the Installation in VSGNet to prepare the V-COCO dataset and install requirements.txt. Note that this is python2 virtual environment. After preparing the dataset by following VSGNet, replace Object_Detections_vcoco folder with our processed data which includes keypoints of humans and objects. Our pretrained model can be downloaded here.

Training

To train the model from scratch (inside "scripts/"):

CUDA_VISIBLE_DEVICES=0, 1 python2 main.py -fw new_test -ba 16 -l 0.005 -e 60 -sa 20 

Test

CUDA_VISIBLE_DEVICES=0, 1 python2 main.py -fw soa_paper -ba 16 -r t -i t

Citing

If you find this work useful, please consider our paper to cite:

@inproceedings{zhu22skeleton,
 author={Zhu, Manli and Ho, Edmond S. L. and Shum, Hubert P. H.},
 booktitle={Proceedings of the 2022 IEEE International Conference on Systems, Man, and Cybernetics},
 series={SMC '22},
 title={A Skeleton-aware Graph Convolutional Network for Human-Object Interaction Detection},
 year={2022},
}

Other

Our SGCN4HOI is based on VSGNet.