Research map / Video / Video-language understanding
YesTrack: Referring Multi-Object Tracking via MLLM-based Yes/No Reasoning
Quansheng Hu, Qin Sun, Qiansen Dai, Jin Ding, Wan Zhang, Xue Zhou, Jianxiao Zou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add video mode (stage 2 refinement) to frame mode (ablation) | YesTrack · Refer-KITTI | HOTA: 52.64 -> 53.47 | Improved |
| add video mode to frame mode (ablation) | YesTrack · Refer-KITTI | inference time: 40mins -> 1h32mins | Hurt |
| add Temporal Confidence Prior (TCP) (ablation) | YesTrack · Refer-KITTI | HOTA: 53.47 -> 54.00 | Improved |
| add TCP (ablation) | YesTrack · Refer-KITTI | AssA: 65.60 -> 66.57 | Improved |
| add Temporal Reference Propagation (TRP) without TCP (ablation) | YesTrack · Refer-KITTI | inference time: 1h32mins -> 22mins | Improved |
| add TRP without TCP (ablation) | YesTrack · Refer-KITTI | AssA: 65.60 -> 65.88 | Improved |
| add TRP on top of TCP model (ablation) | YesTrack · Refer-KITTI | HOTA: 54.00 -> 53.66 | Hurt |
| add TRP on top of TCP model (ablation) | YesTrack · Refer-KITTI | inference time: 1h34mins -> 25mins | Improved |
| replace iKUN referring with YesTrack verifier | two-stage RMOT · Refer-KITTI referential accuracy | accuracy: 84.62 -> 91.15 | Improved |
| replace iKUN referring with YesTrack verifier (ByteTrack tracker) | two-stage RMOT · Refer-KITTI | HOTA: 33.59 -> 44.26 | Improved |
| replace iKUN referring with YesTrack verifier (GT tracks) | two-stage RMOT · Refer-KITTI | HOTA: 55.97 -> 72.80 | Improved |
| replace embedding-based ReID with MLLM pairwise verifier (YesTrack-MOT) | tracking-by-detection · KITTI MOT | HOTA: 41.34 -> 45.10 | Improved |
Closest papers by meaning
- Referring Multi-Object Tracking (CVPR 2023 · similarity 0.92)
- Referring Multi-Object Tracking with Diverse Dynamic Expressions (TPAMI 2026 · similarity 0.92)
- Rethinking Two-Stage Referring-by-Tracking in Referring Multi-Object Tracking: Make it Strong Again (CVPR 2026 · similarity 0.90)
- SOVTrack: Open-Vocabulary Multi-Object Tracking with Self-Supervised Pseudo Labeling and Feature Distillation (ECCV 2026 · similarity 0.90)
- Cross-View Referring Multi-Object Tracking (AAAI 2025 · similarity 0.89)
- ChatTracker: Enhancing Visual Tracking Performance via Chatting with Multimodal Large Language Model (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
