Research map / Detection and segmentation / Tracking
Hypergraph-State Collaborative Reasoning for Multi-Object Tracking
Zikai Song, Junqing Yu, Yi-Ping Phoebe Chen, Wei Yang, Xinchao Wang
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 |
|---|---|---|---|
| replace Kalman filter with HyperSSM motion model (ablation) | tracking pipeline · DanceTrack val | HOTA: 47.0 -> 59.7 | Improved |
| replace Kalman filter with HyperSSM motion model (ablation) | tracking pipeline · MOT17 val-half | HOTA: 64.5 -> 65.5 | Improved |
| add hypergraph collaborative reasoning to SSM (ablation) | SSM motion model · DanceTrack val | HOTA: 53.7 -> 59.7 | Improved |
| add hypergraph collaborative reasoning to SSM (ablation) | SSM motion model · MOT17 val-half | HOTA: 63.4 -> 65.5 | Improved |
| add hypergraph reasoning to SSM at YOLOX-X (ablation) | SSM · MOT17 full train | HOTA: 71.9 -> 74.1 | Improved |
| add hypergraph reasoning to SSM at YOLOX-X (ablation) | SSM · MOT17 full train | fps: 20.3 -> 17.5 | Hurt |
| raise hyperedge threshold theta from 0.5 to 0.8 (ablation) | HyperSSM · MOT17 val-half | HOTA: 47.7 -> 65.5 | Improved |
| raise hyperedge threshold theta from 0.8 to 0.98 (ablation) | HyperSSM · MOT17 val-half | HOTA: 65.5 -> 64.7 | Hurt |
| increase number of HyperSSM layers beyond 8 (ablation) | HyperSSM · MOT17 and DanceTrack | HOTA | No clear effect |
| use 5-frame trajectory segments instead of longer ones (ablation) | HyperSSM · MOT17 and DanceTrack | HOTA | Improved |
| use larger YOLOX detector (S to X) (ablation) | HyperSSM tracker · MOT17 full train | HOTA: 57.3 -> 74.1 | Improved |
| HyperSSM vs best prior on DanceTrack test | HyperSSM · DanceTrack test | HOTA: 65.7 vs 76.2 (DualMOT) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- DiffMOT — DiffMOT: A Real-time Diffusion-based Multiple Object Tracker with Non-linear Prediction (CVPR 2024) · better on 3 of 3 shared results
- HybridSORT — Hybrid-SORT: Weak Cues Matter for Online Multi-Object Tracking (AAAI 2024) · better on 1 of 2 shared results
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
- ReST: A Reconfigurable Spatial-Temporal Graph Model for Multi-Camera Multi-Object Tracking (ICCV 2023 · similarity 0.88)
- Exploiting Hierarchical Dense Structures on Hypergraphs for Multi-Object Tracking (TPAMI 2016 · similarity 0.87)
- Samba: Synchronized Set-of-Sequences Modeling for Multiple Object Tracking (ICLR 2025 · similarity 0.87)
- HyperGLM: HyperGraph for Video Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.86)
- Multi-Scene Generalized Trajectory Global Graph Solver with Composite Nodes for Multiple Object Tracking (AAAI 2024 · similarity 0.86)
- Asynchronous Collaborative Graph Representation for Frames and Events (CVPR 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.
