Research map / Point clouds / Segmentation and detection
Exploring Simple 3D Multi-Object Tracking for Autonomous Driving
Chenxu Luo, Xiaodong Yang, Alan Yuille
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 |
|---|---|---|---|
| use a unified map for tracked and new-born objects instead of separated maps (ablation) | SimTrack (pillar) · nuScenes val | overall AMOTA: 44.6 -> 60.0 | Improved |
| average the current hybrid-time map with the previous updated centerness map (ablation) | SimTrack (pillar) · nuScenes val | overall AMOTA: 50.0 -> 60.0 | Improved |
| combine maps (effect on identity switches) (ablation) | SimTrack (pillar) · nuScenes val | IDS: 4761 -> 1406 | Improved |
| double centerness map resolution from 0.8m to 0.4m (ablation) | SimTrack (pillar) · nuScenes val | overall AMOTA: 60.0 -> 61.1 | Improved |
| double centerness map resolution (pedestrian) (ablation) | SimTrack (pillar) · nuScenes val | pedestrian AMOTA: 68.3 -> 75.0 | Improved |
| double centerness map resolution (motorcycle) (ablation) | SimTrack (pillar) · nuScenes val | motor AMOTA: 57.6 -> 65.9 | Improved |
| joint detection and tracking training vs CenterPoint-based baseline | SimTrack (pillar) · nuScenes val | mAVE: 0.300 -> 0.201 | Improved |
| joint detection and tracking training vs CenterPoint-based baseline | SimTrack (voxel) · nuScenes val | mAVE: 0.272 -> 0.191 | Improved |
| SimTrack tracking instead of CenterPoint matching (same detections) | SimTrack · nuScenes val (pillar) | overall AMOTA: 58.0 -> 60.0 | Improved |
| use SimTrack detections for the Kalman filter baseline | Kalman Filter · nuScenes val (pillar) | overall AMOTA: 51.2 -> 57.4 | Improved |
| loosen CenterPoint max matching distance from 4m to 10m / remove limit (ablation) | CenterPoint · nuScenes val Car | AMOTA: 82.5 -> 59.6 | Hurt |
| do not retain dead tracks (age 0) in CenterPoint (ablation) | CenterPoint · nuScenes val Car | IDS: 238 -> 365 | Hurt |
Closest papers by meaning
- An Effective Motion-Centric Paradigm for 3D Single Object Tracking in Point Clouds (TPAMI 2023 · similarity 0.91)
- Beyond 3D Siamese Tracking: A Motion-Centric Paradigm for 3D Single Object Tracking in Point Clouds (CVPR 2022 · similarity 0.91)
- TrackAny3D: Transferring Pretrained 3D Models for Category-unified 3D Point Cloud Tracking (ICCV 2025 · similarity 0.89)
- Modeling Continuous Motion for 3D Point Cloud Object Tracking (AAAI 2024 · similarity 0.88)
- UAWTrack: Universal 3D Single Object Tracking in Adverse Weather (AAAI 2025 · similarity 0.87)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single 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.
