This repo is the official implementation for Adaptive Hyper-Graph Convolution Network for Skeleton-based Human Action Recognition with Virtual Connections. The paper is accepted to ICCV 2025.

Illustration of Hyper-GCN. In (a), dotted lines represent the virtualconnections. Each coloured part containing multiple joints represents a hyper-graph with hyper-edges. (b) provides an intuitive comparison between normal graph convolution and hyper-graph convolution operations with the same connectivity degree.

The framework of Hyper-GCN. Part (a) represents the Multi-head Hyper-graph Convolution (M-HGC) module. Part (b) represents the process of constructing an adaptive hyper-graph. Part (c) represents the architecture of Hyper-GCN.

Drink water	Stand up	Clapping	Jump up
Neck pain	Wipe face	Hopping	Kicking something

There are some selected actions for visualizing the hyper-graph constructed by Hyper-GCN.

• Python == 3.9
• PyTorch == 1.13.0
• torchpack == 0.2.2
• numpy == 1.26.4
• PyYAML, tqdm, tensorboardX, matplotlib, sklearn, h5py

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

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

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

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

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`
      ...

• 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

• We provide the training configs for base and large version in 4 modalities (joint, bone, joint motion and bone motion) for each benchmark. You can change the config file depending on what you want. As a example for base version on NTU120 X-Sub with device 0:

# train for modality joint
python main.py --config config/base/nturgbd120-cross-subject/hyper_joint.yaml --work-dir <the save path of results> --device 0
# train for modality bone
python main.py --config config/base/nturgbd120-cross-subject/hyper_bone.yaml --work-dir <the save path of results> --device 0
# train for modality joint motion
python main.py --config config/base/nturgbd120-cross-subject/hyper_joint_motion.yaml --work-dir <the save path of results> --device 0
# train for modality bone motion
python main.py --config config/base/nturgbd120-cross-subject/hyper_bone_motion.yaml --work-dir <the save path of results> --device 0

• You can test the trained models saved in <the save path of results> as the follow command:

python main.py --config <the save path of results>/config.yaml --work-dir <the save path of results> --phase test --save-score True --weights <the save path of results>/xxx.pt --device 0

• If you want to ensemble the results of 4 modalities for the final result, run the following command (As a example for base version on NTU120 X-Sub):

python ensemble.py --datasets ntu120/xsub --model base --joint-dir <the save path of result on joint> --bone-dir <the save path of result on bone> --joint-motion-dir <the save path of result on joint motion> --bone-motion-dir <the save path of result on bone motion>

We provide the pretrained weights, configs and training logs. You can download pretrained models for producing the final results on NTU 60 & 120 in Google Drive.

• This repo is based on CTR-GCN.

• The data processing is borrowed from SGN and HCN.

• The training strategy is referenced from BlockGCN, SkateFormer and PYSKL

Thanks to the original authors for their work!

For any questions, feel free to contact: youwei_zhou@stu.jiangnan.edu.cn