HKBU-Mocap Extended dataset

Introduction

HKBU-Mocap Extended is the extended version of our previously released dataset, HKBU-Mocap, designed for research on multi-view, multi-person motion analysis, as presented in our journal paper: Every Angle Is Worth A Second Glance: Mining Kinematic Skeletal Structures from Multi-view Joint Cloud. This dataset contains 3,600 frames (originally recorded at 30 FPS and downsampled to 600 frames for efficient storage) across 5 camera views, capturing performances involving 6 individuals. The dataset features group dance routines that include complex movements such as self-rotations and frequent positional changes. Owing to the dynamic nature of these dance sequences, the diversity and richness of actions in HKBU-Mocap Extend significantly surpass those in our previous release.

All poses are manually annotated using the BODY25 format defined by OpenPose. Since the manual annotation is hard work, some of them are still not perfect. We will continue to maintain the dataset and provide more idealistic data.

Browse

Six actors perform dance routines captured by five cameras. The sequences feature highly dynamic movements and frequent severe occlusions. #multi-person #multi-view

Download link: Google Drive

Specs:

Resolution: 3840 x 2160

Video length: 2 minutes

Annotated frame number: 900

Video Encode CRF: 30

File Structure:

Annotation: coop_8_3D.npy

Video Sequences: video.zip

Frame RGB images: image.zip

Camera parameters: camera_params.npy

Citation

@article{jiang2025every,
  title={Every Angle Is Worth A Second Glance: Mining Kinematic Skeletal Structures from Multi-view Joint Cloud},
  author={Jiang, Junkun and Chen, Jie and Au, Ho Yin and Chen, Mingyuan and Xue, Wei and Guo, Yike},
  journal={IEEE Transactions on Visualization and Computer Graphics},
  year={2025},
  publisher={IEEE}
}