google-research-datasets/LaGOT

We enrich the LaSOT validation set with annotations of additional object tracks, up to 10 object tracks per video in total. Tracks consist of precise bounding box annotations of moving objects. Annotations are provided at 10 fps. The original LaSOT validation set annotations and video can be downloaded from: https://vision.cs.stonybrook.edu/~lasot/

LaGOT

We release the LaGOT evaluation benchmark for both single-object trackers (SOT) and multi-object trackers (MOT) / open-world trackers (OWT).

The LaGOT benchmark is built on top of the LaSOT validation set. We added up to 9 additional tracks per video to the LaSOT validation set annotations.

Download

Additional annotations are located in data/. To obtain pixels, please follow the instructions of the official LaSOT repository.

In addition to the new annotations, we also release the inference results of multiple state-of-the-art SOT and MOT / OWT trackers.

Usage

The scripts for results visualizations are located in scripts/. To use those the pytracking toolbox and/or the TrackEval toolbox is needed.

Citation

If you use LaGOT, please cite our paper:

@InProceedings{Mayer_2024_WACV,
    author    = {Mayer, Christoph and Danelljan, Martin and Yang, Ming-Hsuan and Ferrari, Vittorio and Van Gool, Luc and Kuznetsova, Alina},
    title     = {Beyond SOT: Tracking Multiple Generic Objects at Once},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2024},
    pages     = {6826-6836}
}

Contacts

If you have any questions, please contact lagot-benchmark@googlegroups.com.