Home

Awesome

Hi4D: 4D Instance Segmentation of Close Human Interaction

Project Page | Paper | Video | Data

<img src="assets/teaser.png"/> Hi4D (Humans interacting in 4D), a dataset of humans in close physical interaction that contains (A) 4D textured scans, (B) instance meshes with vertex-level contact annotations, (C) instance segmentation masks in 2D and 3D, (D) registered parametric body models with contact annotations.

Contents

  1. Dataset
  2. Visualization
  3. Citation
  4. Contact

Hi4D Dataset

Please fill out the Hi4D Application Form to access Hi4D. We will send you an email with more information after approval of your application.

Hi4D dataset structure can be found in Dataset Structure.

Visualization

We use AITViewer to visualize the dataset.

Installation

      pip install aitviewer

Usage

  1. Change the smplx_models parameter in visualize_hi4d.py to the path of SMPL models. More information please refer to AITViewer Documentation.
  2. Change the HI4D_PATH parameter in visualize_hi4d.py to the path of Hi4D dataset.
  3. Run the following command to visualize the dataset.
      python visualize_hi4d.py --pair {PAIRXX} --action {ACTIONXX} --vis {VIS_TYPE}

Citation

If you find this work useful for your research, please cite our paper:

@inproceedings{yin2023hi4d,
      author = {Yin, Yifei and Guo, Chen and Kaufmann, Manuel and Zarate, Juan and Song, Jie and Hilliges, Otmar}, 
      title = {Hi4D: 4D Instance Segmentation of Close Human Interaction}, 
      booktitle = {Computer Vision and Pattern Recognition (CVPR)},
      year = {2023}
      }

Contact

If you have any questions, please contact Chen Guo.