Research map / Detection and segmentation / Tracking
Referring Multi-Object Tracking
Dongming Wu, Wencheng Han, Tiancai Wang, Xingping Dong, Xiangyu Zhang, Jianbing Shen
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 |
|---|---|---|---|
| use TransRMOT instead of the strongest reported baseline, TrackFormer | Refer-KITTI tracking · Refer-KITTI | HOTA: 33.26 -> 35.54 | Improved |
| add the cross-modal fusion module (ablation) | TransRMOT · Refer-KITTI | HOTA: 17.01 -> 35.54 | Improved |
| use early fusion instead of concatenating visual and linguistic features (ablation) | TransRMOT · Refer-KITTI | HOTA: 28.61 -> 35.54 | Improved |
| use early fusion instead of language as decoder query (ablation) | TransRMOT · Refer-KITTI | HOTA: 33.29 -> 35.54 | Improved |
| add track queries for cross-frame association (ablation) | TransRMOT · Refer-KITTI | HOTA | Improved |
| use track queries instead of SORT association (ablation) | TransRMOT · Refer-KITTI | HOTA: 31.63 -> 35.54 | Improved |
| use track queries instead of ByteTrack association (ablation) | TransRMOT · Refer-KITTI | HOTA: 32.12 -> 35.54 | Improved |
| replace RoBERTa with FastText or GloVe (ablation) | TransRMOT · Refer-KITTI | HOTA: 32.39, 32.45 vs 35.54 | Hurt |
| replace RoBERTa with BERT or Distill-BERT (ablation) | TransRMOT · Refer-KITTI | HOTA: 35.28, 33.56 vs 35.54 | No clear effect |
| vary the referring threshold from 0.2 to 0.5 (ablation) | TransRMOT · Refer-KITTI | HOTA: 35.10, 35.07, 35.54, 34.73 | No clear effect |
| increase the referring threshold to 0.6 or 0.7 (ablation) | TransRMOT · Refer-KITTI | HOTA: 35.54 -> 31.09; 31.63 | Hurt |
| train with 50% or 75% of expressions instead of all expressions (ablation) | TransRMOT · Refer-KITTI | HOTA: 35.54 -> 35.19; 35.48 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ByteTrack — ByteTrack: Multi-Object Tracking by Associating Every Detection Box (ECCV 2022) · better on 1 of 1 shared result
- FairMOT — FairMOT: On the Fairness of Detection and Re-identification in Multiple Object Tracking (IJCV 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Referring Multi-Object Tracking with Diverse Dynamic Expressions (TPAMI 2026 · similarity 0.95)
- Rethinking Two-Stage Referring-by-Tracking in Referring Multi-Object Tracking: Make it Strong Again (CVPR 2026 · similarity 0.92)
- YesTrack: Referring Multi-Object Tracking via MLLM-based Yes/No Reasoning (ECCV 2026 · similarity 0.92)
- Cross-View Referring Multi-Object Tracking (AAAI 2025 · similarity 0.89)
- HFF-Tracker: A Hierarchical Fine-grained Fusion Tracker for Referring Multi-Object Tracking (AAAI 2025 · similarity 0.88)
- iKUN: Speak to Trackers without Retraining (CVPR 2024 · similarity 0.87)
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.
