Research map / Detection and segmentation / Tracking
Towards Real-Time Multi-Object Tracking
Zhongdao Wang, Liang Zheng, Yixuan Liu, Yali Li, Shengjin 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 |
|---|---|---|---|
| train embedding with cross-entropy instead of triplet loss (ablation) | JDE · Embedding validation | TPR@FAR=0.1: 44.3 -> 88.2 | Improved |
| train embedding with cross-entropy instead of triplet loss (ablation) | JDE · Caltech val / MOT-15 train | MOTA: 59.5 -> 64.3 | Improved |
| use triplet upper bound instead of plain triplet loss (ablation) | JDE · Embedding validation | TPR@FAR=0.1: 42.2 -> 44.3 | Improved |
| use uncertainty weighting instead of approximate-optimal searched weights (ablation) | JDE · Caltech val / MOT-15 train | MOTA: 64.3 -> 65.8 | Improved |
| use uncertainty weighting instead of uniform loss weights (ablation) | JDE · Caltech val (detection AP) | AP: 6.8 -> 83.0 | Improved |
| use uncertainty weighting instead of uniform loss weights (ablation) | JDE · Caltech val / MOT-15 train | MOTA: 36.9 -> 65.8 | Improved |
| use MGDA-UB weighting instead of uncertainty weighting (ablation) | JDE · Caltech val / MOT-15 train | MOTA: 65.8 -> 38.3 | Hurt |
| use loss normalization instead of uncertainty weighting (ablation) | JDE · Caltech val / MOT-15 train | MOTA: 65.8 -> 57.9 | Hurt |
| use JDE online association instead of SORT cascade matching (ablation) | JDE · MOT-15 (low density) | MOTA: 66.9 -> 67.5 | Improved |
| use JDE online association instead of SORT cascade matching (ablation) | JDE · CVPR19-01 (high density) | FPS: 26.4 -> 33.9 | Improved |
| use JDE online association instead of SORT (ablation) | JDE · MOT-15 (low density) | IDF-1: 55.8 -> 67.6 | Improved |
| reduce input resolution from 1088x608 to 864x408 | JDE · MOT-16 test | MOTA: 64.4 -> 62.1 | Hurt |
Closest papers by meaning
- Improving Multiple Object Tracking With Single Object Tracking (CVPR 2021 · similarity 0.90)
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.90)
- From Detection to Association: Learning Discriminative Object Embeddings for Multi-Object Tracking (CVPR 2026 · similarity 0.90)
- More Than Meets the Eye: Enhancing Multi-Object Tracking Even with Prolonged Occlusions (ICML 2025 · similarity 0.89)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.88)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · similarity 0.88)
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.
