Research map / Detection and segmentation / Tracking
Multi-Object Tracking and Segmentation Via Neural Message Passing
Guillem Brasó, Orcun Cetintas, Laura Leal-Taixé
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 time-aware node update instead of vanilla node update (ablation) | MPNTrack · MOT17 (3-fold CV) | IDF1: 67.1 -> 70.0 | Improved |
| add time-aware node update (ablation) | MPNTrack · MOT17 (3-fold CV) | constraint satisfaction (%): 83.2 -> 98.8 | Improved |
| add time-aware node update (ablation) | MPNTrack · MOT17 (3-fold CV) | ID switches: 917 -> 602 | Improved |
| add time-aware node update (ablation) | MPNTrack · MOT17 (3-fold CV) | FP: 3239 -> 6169 | Hurt |
| add relative position edge features to time features (ablation) | MPNTrack · MOT17 edge features | IDF1: 52.6 -> 68.7 | Improved |
| add CNN embedding distance to time+position edge features (ablation) | MPNTrack · MOT17 edge features | ID switches: 895 -> 602 | Improved |
| add CNN embedding distance to time+position edge features (ablation) | MPNTrack · MOT17 edge features | MOTA: 63.6 -> 64.0 | Improved |
| increase message passing steps from 0 to 2 (ablation) | MPNTrack · MOT17 (3-fold CV) | IDF1 and MOTA | Improved |
| increase message passing steps beyond twelve (ablation) | MPNTrack · MOT17 (3-fold CV) | IDF1 and MOTA | No clear effect |
| use attentive instead of additive message passing for mask features (ablation) | MPNSeg · MOTS20 (4-fold CV) | sMOTSA: 57.3 -> 60.7 | Improved |
| add raw node features skip connection to attentive updates (ablation) | MPNSeg · MOTS20 (4-fold CV) | IDF1: 72.0 -> 73.1 | Mixed |
| add attentive message passing for mask features over raw-only features (ablation) | MPNSeg · MOTS20 (4-fold CV) | sMOTSA: 60.5 -> 60.7 | No clear effect |
Closest papers by meaning
- Learning a Neural Solver for Multiple Object Tracking (CVPR 2020 · similarity 0.94)
- MOTS: Multi-Object Tracking and Segmentation (CVPR 2019 · similarity 0.91)
- Learning Multi-Object Tracking and Segmentation From Automatic Annotations (CVPR 2020 · similarity 0.90)
- Learnable Graph Matching: Incorporating Graph Partitioning With Deep Feature Learning for Multiple Object Tracking (CVPR 2021 · similarity 0.89)
- Multi-Scene Generalized Trajectory Global Graph Solver with Composite Nodes for Multiple Object Tracking (AAAI 2024 · similarity 0.88)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 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.
