Research map / Detection and segmentation / 3D detection and driving
Center-Based 3D Object Detection and Tracking
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace anchor-based detection with center-based detection using VoxelNet (ablation) | 3D detection · Waymo validation Level 2 | mAPH: 60.3 -> 64.6 | Helped |
| replace anchor-based detection with center-based detection using PointPillars (ablation) | 3D detection · Waymo validation Level 2 | mAPH: 57.5 -> 62.0 | Helped |
| replace anchor-based detection with center-based detection using VoxelNet (ablation) | 3D detection · nuScenes validation | mAP: 52.6 -> 56.4 | Helped |
| add surface-center point features to second-stage refinement (ablation) | CenterPoint-Voxel · Waymo validation Level 2 vehicle | mAPH: 66.5 -> 68.3 | Helped |
| add surface-center point features to second-stage refinement (ablation) | CenterPoint-Voxel · Waymo validation Level 2 pedestrian | mAPH: 62.7 -> 65.3 | Helped |
| add second-stage refinement (ablation) | CenterPoint · nuScenes | detection performance | No clear effect |
| add Voxel Set Abstraction features (ablation) | CenterPoint second-stage refinement · Waymo validation Level 2 pedestrian | mAPH: 65.3 -> 65.2 | Hurt |
| add radial basis function interpolation of voxel features (ablation) | CenterPoint second-stage refinement · Waymo validation Level 2 vehicle | mAPH: 68.3 -> 68.4 | Helped |
| use learned velocity and closest-point matching instead of a Mahalanobis Kalman filter (ablation) | CenterPoint-Voxel tracker · nuScenes validation | AMOTA: 60.0 -> 63.7 | Helped |
| add image-based PointPainting features (ablation) | CenterPoint · nuScenes validation | NDS: 65.4 -> 68.0 | Helped |
| add flip and rotation test-time augmentation (ablation) | CenterPoint · nuScenes validation | NDS: 68.0 -> 70.3 | Helped |
| ensemble five models (ablation) | CenterPoint · nuScenes validation | NDS: 70.3 -> 71.4 | Helped |
Closest papers by meaning
- Generalizable Structure-Aware Keypoint Correspondence for Category-Unified 3D Single Object Tracking (CVPR 2026 · similarity 0.88)
- Beyond 3D Siamese Tracking: A Motion-Centric Paradigm for 3D Single Object Tracking in Point Clouds (CVPR 2022 · similarity 0.87)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · similarity 0.87)
- CXTrack: Improving 3D Point Cloud Tracking With Contextual Information (CVPR 2023 · similarity 0.86)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.85)
- SPGroup3D: Superpoint Grouping Network for Indoor 3D Object Detection (AAAI 2024 · similarity 0.85)
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.