Object Re-identification from Point Clouds

Datasets

As part of the paper "Object Re-identification from Point Clouds" published at WACV'24 [1], we release two datasets for object re-identification from point-clouds and images: nuScenes reid and Waymo reid. Each dataset contains object observations cropped from images or LiDAR scans in different scenes. While nuScenes reid is fully open source and available for download at the GitHub repository [2], accessing the Waymo reid dataset requires agreeing to the Waymo Open Dataset (WOD) [4] licence. Please send me an email at benjtherien at gmail dot com to access it.

References

  1. Thérien, Benjamin, "Object Re-Identification from Point Clouds", Winter Conference on Applications of Computer Vision (WACV), 2024.
  2. Thérien, Benjamin, "Object Re-identification from Point Clouds", GitHub repository, 2024.
  3. Thérien, Benjamin, "Towards Object Re-identification from Point Clouds for 3D MOT", MMath thesis, 2023.
  4. Waymo, "Waymo Open Dataset: An autonomous driving dataset", 2019.