Research map / Detection and segmentation / 3D detection and driving
Monocular Quasi-Dense 3D Object Tracking
Hou-Ning Hu, Yung-Hsu Yang, Tobias Fischer, Trevor Darrell, Fisher Yu, Min Sun
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 deep feature affinity (ablation) | QD3DT KF3D · nuScenes val | AMOTA@1: 0.1779 -> 0.2306 | Improved |
| add 3D IoU affinity (ablation) | QD3DT KF3D · nuScenes val | AMOTA@1: 0.2013 -> 0.2306 | Improved |
| add depth-order affinity (ablation) | QD3DT KF3D · nuScenes val | AMOTA@1: 0.2213 -> 0.2306 | Improved |
| add depth center term (ablation) | QD3DT KF3D · nuScenes val | AMOTA@1: 0.2301 -> 0.2306 | No clear effect |
| use full frames at inference with VeloLSTM (ablation) | VeloLSTM · nuScenes val | AMOTA@1: 0.2329 -> 0.2421 | Improved |
| use full frames at inference with KF3D (ablation) | KF3D · nuScenes val | AMOTA@1: 0.2306 -> 0.2321 | Improved |
| use 3D IoU instead of 2D IoU in A_iou (ablation) | QD3DT · nuScenes val | AMOTA@1: 0.1381 -> 0.2306 | Improved |
| use 3D IoU instead of BEV IoU in A_iou (ablation) | QD3DT · nuScenes val | AMOTA@1: 0.1929 -> 0.2306 | Improved |
| use motion-aware association instead of cosine (ablation) | QD3DT · nuScenes val | AMOTA@1: 0.2234 -> 0.2306 | Improved |
| use motion-aware association instead of centroid (ablation) | QD3DT · nuScenes val | AMOTA@1: 0.2300 -> 0.2306 | No clear effect |
| add VeloLSTM motion model instead of detection only (ablation) | QD3DT · nuScenes val | AMOTA@1: 0.222 -> 0.242 | Improved |
| use VeloLSTM instead of KF3D (ablation) | QD3DT · nuScenes val | AMOTA@1: 0.232 -> 0.242 | Improved |
Closest papers by meaning
- Joint Monocular 3D Vehicle Detection and Tracking (ICCV 2019 · similarity 0.92)
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.88)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.87)
- TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels (NeurIPS 2025 · similarity 0.87)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.87)
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 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.
