Track to Detect and Segment: An Online Multi-Object Tracker
About
Most online multi-object trackers perform object detection stand-alone in a neural net without any input from tracking. In this paper, we present a new online joint detection and tracking model, TraDeS (TRAck to DEtect and Segment), exploiting tracking clues to assist detection end-to-end. TraDeS infers object tracking offset by a cost volume, which is used to propagate previous object features for improving current object detection and segmentation. Effectiveness and superiority of TraDeS are shown on 4 datasets, including MOT (2D tracking), nuScenes (3D tracking), MOTS and Youtube-VIS (instance segmentation tracking). Project page: https://jialianwu.com/projects/TraDeS.html.
Jialian Wu, Jiale Cao, Liangchen Song, Yu Wang, Ming Yang, Junsong Yuan• 2021
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multiple Object Tracking | MOT17 (test) | MOTA69.1 | 921 | |
| Video Instance Segmentation | YouTube-VIS 2019 (val) | AP32.6 | 567 | |
| Multi-Object Tracking | DanceTrack (test) | HOTA0.48 | 355 | |
| Multi-Object Tracking | MOT16 (test) | MOTA70.1 | 228 | |
| Video Instance Segmentation | YouTube-VIS (val) | AP32.6 | 118 | |
| 3D Multi-Object Tracking | nuScenes (val) | AMOTA11.8 | 115 | |
| Multi-Object Tracking | MOT 2016 (test) | MOTA70.1 | 59 | |
| Multi-Object Tracking | MOT17 1.0 (test) | MOTA69.1 | 48 | |
| Multi-Object Tracking | MOT 2017 (test) | MOTA69.1 | 34 | |
| 3D Object Tracking | nuScenes (test) | -- | 28 |
Showing 10 of 15 rows