TAO Multi-target Tracking Dataset

TAO stands for Tracking Any Object Dataset, which is a joint dataset for tracking any object.

The TAO dataset consists of 2,907 high-definition videos of different environments, with an average length of 36.8 seconds and 833 categories. Compared with existing target tracking datasets, the TAO dataset has richer sample types.

The dataset is annotated by marking the trajectories of all moving objects in the video and naming them afterwards.