RMOT26 is a large-scale benchmark for Query-Driven Multi-Object Tracking, introduced in the paper QTrack: Query-Driven Reasoning for Multi-modal MOT.
Multi-object tracking (MOT) has traditionally focused on estimating trajectories of all objects in a video. RMOT26 introduces a query-driven tracking paradigm that… See the full description on the dataset page:
https://huggingface.co/datasets/GAASH-Lab/RMOT26.