Research map / Detection and segmentation / Tracking
Track Without Appearance: Learn Box and Tracklet Embedding With Local and Global Motion Patterns for Vehicle Tracking
Gaoang Wang, Renshu Gu, Zuozhu Liu, Weijie Hu, Mingli Song, Jenq-Neng Hwang
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 binary cross-entropy loss to triplet loss for box embedding (ablation) | LGM box embedding · KITTI-car tracking validation | HOTA: 75.7 -> 76.9 | Improved |
| add binary cross-entropy loss to triplet loss for box embedding (ablation) | LGM box embedding · KITTI-car tracking validation | MOTA: 88.0 -> 89.0 | Improved |
| add binary cross-entropy loss to tracklet embedding loss (ablation) | LGM tracklet embedding · KITTI-car tracking validation | HOTA: 76.3 -> 76.9 | Improved |
| add binary cross-entropy loss to tracklet embedding loss (ablation) | LGM tracklet embedding · KITTI-car tracking validation | MOTA: 88.2 -> 89.0 | Improved |
| add reconstruction loss to tracklet embedding training (ablation) | LGM tracklet embedding · KITTI-car tracking validation | convergence | Improved |
| add tracklet embedding module to box embedding (ablation) | LGM · KITTI-car tracking validation | HOTA: 75.2 -> 76.9 | Improved |
| add tracklet embedding module to box embedding (ablation) | LGM · KITTI-car tracking validation | AssA: 74.3 -> 77.2 | Improved |
| add tracklet embedding module to box embedding (ablation) | LGM · KITTI-car tracking validation | MOTA: 88.0 -> 89.0 | Improved |
| use LGM motion-only tracker instead of appearance-based CenterTrack | vehicle tracking · KITTI-car tracking test | MOTA: 88.8 -> 87.6 | Hurt |
| use LGM motion-only tracker instead of appearance-based CenterTrack | vehicle tracking · KITTI-car tracking test | HOTA: 73.0 -> 73.1 | No clear effect |
| use LGM instead of best prior method FAMNet | vehicle tracking · UA-Detrac test | PR-MOTA: 19.8 -> 22.5 | Improved |
| use LGM instead of IOU Tracker on pedestrians | motion-only pedestrian tracking · MOT17 pedestrian test | MOTA: 45.5 -> 56.0 | Improved |
Closest papers by meaning
- GNN3DMOT: Graph Neural Network for 3D Multi-Object Tracking With 2D-3D Multi-Feature Learning (CVPR 2020 · similarity 0.88)
- OVTrack: Open-Vocabulary Multiple Object Tracking (CVPR 2023 · similarity 0.88)
- Learnable Graph Matching: Incorporating Graph Partitioning With Deep Feature Learning for Multiple Object Tracking (CVPR 2021 · similarity 0.87)
- Tracking Without Bells and Whistles (ICCV 2019 · similarity 0.87)
- Monocular Quasi-Dense 3D Object Tracking (TPAMI 2022 · similarity 0.87)
- 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.
