Datasets:

License:

You need to agree to share your Hugging Face information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Please read the ZJU-OLAT Non-Commercial Research License at https://huggingface.co/datasets/STLAND/ZJU_OLAT/blob/main/LICENSE.txt before requesting access. The dataset is for non-commercial research only.

Log in or Sign Up to review the conditions and access this dataset content.

ZJU-OLAT Dataset

ZJU-OLAT contains the dynamic JCam subset and the static BMD subset used in DynOLAT: OLAT Video Prediction for Controllable Dynamic Full-Body Relighting.

Project page Β· Paper (PDF) Β· Files and versions

Download and extracted structure

The repository stores cases in TAR archives. Data archives contain processed cases; separate *_check.tar archives contain visual inspection videos. Manifests list the cases in each release and .sha256 files can be used to verify downloads. The remaining_20261019 batches use names such as *_data_01.tar and *_check.tar.

For example, if the data archives for each subset are extracted into their own directories, the result has the following illustrative layout. Existing archives and check videos need not be extracted to these exact paths:

ZJU-OLAT/
β”œβ”€β”€ BMD/
β”‚   β”œβ”€β”€ BMD01_olat_para/
β”‚   β”‚   β”œβ”€β”€ masks/<view>.mp4
β”‚   β”‚   └── ins_0000/
β”‚   β”‚       β”œβ”€β”€ input/<view>.mp4
β”‚   β”‚       β”œβ”€β”€ input_mask/<view>.mp4
β”‚   β”‚       └── re_target/<view>.mp4
β”‚   └── BMD02_olat_para/ ...
└── Jcam/
    β”œβ”€β”€ calibration/
    β”‚   β”œβ”€β”€ intri.yml
    β”‚   └── extri.yml
    β”œβ”€β”€ tall/<case>_olat_para/ ...
    β”œβ”€β”€ grand/<case>_olat_para/ ...
    └── venti/<case>_olat_para/
        β”œβ”€β”€ input/<view>/<olat>.mp4
        β”œβ”€β”€ re_target/<view>/<olat>.mp4
        β”œβ”€β”€ masks_video/<view>.mp4
        β”œβ”€β”€ masks_video_index/<view>.json
        β”œβ”€β”€ bboxes/<view>.json
        β”œβ”€β”€ flf/input/<view>/start.mp4
        β”œβ”€β”€ flf/input/<view>/end.mp4
        β”œβ”€β”€ flf/bboxes/<view>.json
        β”œβ”€β”€ rsNet_flat_detection/<view>.json
        β”œβ”€β”€ rsNet_invalid_views.json
        └── *_done

In JCam, <view> identifies a camera, and <olat> identifies an OLAT segment. input and re_target are spatially matched video pairs; masks_video holds foreground masks and masks_video_index maps mask frames to source frame IDs. bboxes contains foreground crops, while flf contains the start and end flat-lit clips. JCam calibration is provided separately in Jcam/calibration/. The cropped videos do not use the native calibration directly; see the data guide for spatial and temporal alignment.

For the original archive layout, extraction details, calibration, and helper scripts, see JCam data layout and the BMD release notes. Check each archive's manifest for its actual contents and the associated check videos.

Citation

If you use this dataset, please cite the DynOLAT paper:

@article{liu2026dynolat,
  author = {Liu, Taoming and ShenTu, Yinji and He, Guangzhao and Jin, Yudong and Ye, Yukang and Zhang, Siyu and Jia, Fan and Hua, Wei and Bao, Hujun and Zhou, Xiaowei},
  title = {{DynOLAT}: {OLAT} Video Prediction for Controllable Dynamic Full-Body Relighting},
  journal = {ACM Transactions on Graphics},
  number = {6},
  volume = {45},
  month = {December},
  year = {2026}
}

The repository also provides a separate dataset citation.

License

The ZJU-OLAT dataset, including both JCam and BMD, is available for non-commercial research purposes only. Commercial use is prohibited. See LICENSE.txt for the full terms.

Downloads last month
1,479