OccTrack360

OccTrack360 provides autonomous-driving data organized in the KITTI-360 directory layout, including perspective and fisheye images, point clouds, calibration, poses, and labels.

Directory structure

calibration/
data_2d_raw/
data_3d_raw/
data_poses/
fisheye/
labels/
kitti_360_match.txt

Data files are stored in their corresponding modality and sequence directories. Some large image and point-cloud directories contain numbered subdirectories such as __hf_part_00001 to respect the Hub's per-directory file limit. The inventory release_inventory.json.gz records stored paths, original paths, and byte sizes.

Download and restore

pip install -U huggingface_hub
hf download YouthZestLin/OccTrack360 --repo-type dataset --local-dir OccTrack360
python OccTrack360/restore_hf_dataset.py --check
python OccTrack360/restore_hf_dataset.py

The restore script requires Python 3.8+ and no third-party dependencies. It defaults to the directory containing the script; use --root /path/to/OccTrack360 to select another location. It checks files before moving them and refuses destination conflicts. Calibration, poses, labels, and other files already at their original paths remain in place. ZIP archives remain as archives.

Restoring puts files back into their original logical directories. Those local directories can contain more than 10,000 files, which is expected.

Source and license

KITTI-360 source data is credited to Yiyi Liao, Jun Xie, and Andreas Geiger. The official project page publishes KITTI-360 under CC BY-NC-SA 3.0. Preserve the original attribution, noncommercial and share-alike conditions, and follow the source project's registration and use requirements. Third-party annotations remain subject to their respective original terms.

Please cite: Yiyi Liao, Jun Xie, and Andreas Geiger. "KITTI-360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D." Pattern Analysis and Machine Intelligence (PAMI), 2022.

中文说明

数据按图像、点云、标定、位姿和标签分别存放在对应目录。为适应 Hub 的目录文件数量限制，部分目录采用编号子目录存储。下载后先执行 restore_hf_dataset.py --check，再执行 restore_hf_dataset.py，即可将编号子目录中的文件恢复到原始逻辑目录。数据使用须遵循原始来源的许可条件。