ECCV 2020 TAO Challenge

TAO is a federated dataset for Tracking Any Object, containing 2,907 high resolution videos, captured in diverse environments, which are half a minute long on average. We adopt a bottom-up approach for discovering a large vocabulary of 833 categories, an order of magnitude more than prior tracking benchmarks. This challenge evaluates multi-object tracking on the 488 categories in TAO which overlap with the LVIS (v0.5) dataset. For questions regarding this challenge, please email tao@motchallenge.net.

TAO ECCV challenge deadline: August 16th at 11:59PM Anywhere On Earth (AOE, UTC-12)

Training Set

Sample Name FPS Resolution Length Tracks BoxesDensityDescriptionSourceRef.
TAO_val301280x7201460666 (31:29)54851131120.1Validation set, for tuning hyperparameterslink[1]
TAO_train301280x720764526 (04:44)2647546390.1Training set, for training trackerslink[1]
Total 2225192 frm.
(74173 s.)
8132 167751 0.1

Test Set

Sample Name FPS Resolution Length Tracks BoxesDensityDescriptionSourceRef.
TAO_test301280x7202221846 (34:22)79721646500.1Test set, for benchmarkinglink[1]
Total 2221846 frm.
(74062 s.)
7972 164650 0.1


Download

Get all data (226GB + 121GB)
Get files (no img) only (6.4MB)
Development Kit

References:


[1] Dave, A., Khurana, T., Tokmakov, P., Schmid, C. & Ramanan, D. TAO: A Large-Scale Benchmark for Tracking Any Object. In European Conference on Computer Vision, 2020.