Research map / Detection and segmentation / Tracking
Unifying Short and Long-Term Tracking With Graph Hierarchies
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Hierarchy of nine SUSHI blocks versus a single level (monolithic graph) with increasing clip length (ablation) | SUSHI tracker · MOT17 train (ablation) | IDF1: +13.4 IDF1 over naive single-level baseline | Helped |
| Increasing clip length in a single non-hierarchical level (ablation) | single-level GNN · MOT17 train (ablation) | IDF1 | Hurt |
| Increasing the number of hierarchy levels up to nine for a fixed clip size (ablation) | SUSHI tracker · MOT17 train (ablation) | IDF1 | Helped |
| Sharing GNN weights across hierarchy levels with level embeddings | SUSHI tracker · MOT17 train (ablation) | convergence and performance | Helped |
| Adding appearance (reID) edge features at levels 7-9 (ablation) | SUSHI tracker · MOT17 train (ablation) | IDF1: -3.9 IDF1 when disabled for levels 7-9 | Helped |
| Adding position features only at short-term levels (ablation) | SUSHI tracker · MOT17 train (ablation) | IDF1 | No clear effect |
| Adding motion edge features across levels (ablation) | SUSHI tracker · MOT17 train (ablation) | IDF1 | Helped |
| Using the hierarchy with GNN at all levels in MOT17 public setting versus prior graph methods | SUSHI tracker · MOT17 Public | IDF1: +9.8 IDF1 over MPNTrack | Helped |
| Hierarchical unified approach on MOT20 public | SUSHI tracker · MOT20 Public | IDF1: +9.1 IDF1 over prior state of the art | Helped |
| Hierarchy with unified GNN on DanceTrack | SUSHI tracker · DanceTrack | IDF1: +9.5 IDF1 over state of the art | Helped |
| Hierarchical GNN applied naively to all BDD classes | SUSHI tracker · BDD test | mIDF1: +4.2 mIDF1 over state of the art | Helped |
| Naively extending MPNTrack++ single-level GNN to longer horizons (150 and 512 frames) (ablation) | MPNTrack++ · MOT17 train (ablation) | IDF1 | No clear effect |
Closest papers by meaning
- Multi-Scene Generalized Trajectory Global Graph Solver with Composite Nodes for Multiple Object Tracking (AAAI 2024 · similarity 0.89)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.89)
- Learnable Graph Matching: Incorporating Graph Partitioning With Deep Feature Learning for Multiple Object Tracking (CVPR 2021 · similarity 0.89)
- Samba: Synchronized Set-of-Sequences Modeling for Multiple Object Tracking (ICLR 2025 · similarity 0.87)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 2024 · similarity 0.86)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.86)
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.