Research map / Detection and segmentation / Tracking
Rethinking Two-Stage Referring-by-Tracking in Referring Multi-Object Tracking: Make it Strong Again
Weize Li, Yunhao Du, Qixiang Yin, Zhicheng Zhao, Fei Su
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 C-Hook sampling-based feature construction to iKUN-style pipeline (ablation) | FlexHook-iKUN · Refer-KITTI-V2 | HOTA: 10.32 -> 34.49 | Improved |
| replace cosine similarity with Pairwise Correspondence Decoder (ablation) | FlexHook-iKUN · Refer-KITTI-V2 | HOTA: 34.49 -> 38.62 | Improved |
| add Temporal Integration with grid displacement instead of pooling (ablation) | FlexHook-iKUN · Refer-KITTI-V2 | HOTA: 38.62 -> 39.19 | Improved |
| add Conditioning Enhancement with M=10 reference points (ablation) | FlexHook-best · Refer-KITTI-V2 | HOTA: 41.94 -> 42.53 | Improved |
| add Conditioning Enhancement with M=10 (ablation) | FlexHook-CLIP · Refer-KITTI-V2 | HOTA: 40.59 -> 41.42 | Improved |
| increase reference points from 10 to 30 (ablation) | FlexHook-best · Refer-KITTI-V2 | HOTA: 42.53 -> 41.56 | Hurt |
| add noise augmentation in Neighboring Grid Sampling (ablation) | FlexHook-best (no enhancement) · Refer-KITTI-V2 | HOTA: 40.65 -> 41.94 | Improved |
| freeze visual backbone and text encoder (ablation) | FlexHook-best · Refer-KITTI-V2 | HOTA: 42.53 -> 40.86 | Hurt |
| freeze text encoder only (ablation) | FlexHook-best · Refer-KITTI-V2 | HOTA: 42.53 -> 40.84 | Hurt |
| freeze visual backbone only (ablation) | FlexHook-best · Refer-KITTI-V2 | HOTA: 42.53 -> 41.83 | Hurt |
| freeze encoders in CLIP-based FlexHook (ablation) | FlexHook-CLIP · Refer-KITTI-V2 | HOTA: 41.42 -> 40.95 | No clear effect |
| use weaker detector and tracker (D-DETR and StrongSORT) | FlexHook-best · Refer-KITTI-V2 | HOTA: 42.53 -> 40.73 | Hurt |
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.
- iKUN — iKUN: Speak to Trackers without Retraining (CVPR 2024) · better on 2 of 2 shared results
- BYTETrack — ByteTrack: Multi-Object Tracking by Associating Every Detection Box (ECCV 2022) · 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
- Cross-View Referring Multi-Object Tracking (AAAI 2025 · similarity 0.86)
- iKUN: Speak to Trackers without Retraining (CVPR 2024 · similarity 0.86)
- HFF-Tracker: A Hierarchical Fine-grained Fusion Tracker for Referring Multi-Object Tracking (AAAI 2025 · similarity 0.86)
- A Large-Scale Human-Centric Benchmark for Referring Expression Comprehension in the LMM Era (NeurIPS 2024 · similarity 0.85)
- RefDetector: A Simple Yet Effective Matching-based Method for Referring Expression Comprehension (AAAI 2025 · similarity 0.85)
- OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer (ICLR 2025 · similarity 0.85)
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.
