Research map / Detection and segmentation / 3D detection and driving
Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection
Read the paperCheck 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 |
|---|---|---|---|
| Add 3D voxel RoI refinement with a detect head to the BEV one-stage detector (ablation) | Voxel R-CNN · KITTI val, Car | 3D AP Moderate: 77.99 -> 84.21 | Helped |
| Add a detect head using ball query and original PointNet aggregation (ablation) | BEV one-stage baseline · KITTI val, Car | FPS: 40.8 -> 17.4 | Hurt |
| Replace ball query with voxel query (ablation) | Voxel R-CNN detect head · KITTI val, Car | FPS: 17.4 -> 19.1 | Helped |
| Use accelerated PointNet aggregation instead of original PointNet (ablation) | Voxel R-CNN detect head · KITTI val, Car | FPS: 17.4 -> 21.4 | Helped |
| Combine detect head, voxel query, and accelerated PointNet aggregation (ablation) | Voxel R-CNN · KITTI val, Car | FPS: 40.8 -> 25.2 | Helped |
| Use voxel-only two-stage refinement rather than point-based PV-RCNN | Voxel R-CNN · KITTI test, Car, Moderate | 3D AP: 81.43 -> 81.62 | Helped |
| Use voxel-only two-stage refinement rather than point-based PV-RCNN | Voxel R-CNN · Waymo validation, LEVEL_1 vehicle | 3D mAP: 70.30 -> 75.59 | Helped |
| Use voxel-only two-stage refinement rather than point-based PV-RCNN | Voxel R-CNN · Waymo validation, LEVEL_1 vehicle, 50m-Inf | 3D mAP: +10.98% mAP | Helped |
| Use 3D voxel feature volumes for RoI refinement rather than BEV features alone (ablation) | Voxel R-CNN · KITTI val, Car | 3D AP Moderate: 77.99 -> 84.52 | Helped |
Closest papers by meaning
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.93)
- VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking (CVPR 2023 · similarity 0.92)
- PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020 · similarity 0.91)
- Voxel Mamba: Group-Free State Space Models for Point Cloud based 3D Object Detection (NeurIPS 2024 · similarity 0.89)
- GO-N3RDet: Geometry Optimized NeRF-enhanced 3D Object Detector (CVPR 2025 · similarity 0.88)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (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.