Research map / Point clouds / Segmentation and detection
PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection
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 |
|---|---|---|---|
| Integrate voxel CNN features with point-based keypoint abstraction and RoI-grid pooling | PV-RCNN · KITTI test, Car 3D detection | mAP: +1.58, +1.72, +1.73 | Helped |
| Use PV-RCNN rather than prior methods | Waymo Open v1.0 vehicle LEVEL 1 · Waymo Open validation | 3D mAP: +7.37 | Helped |
| Add voxel-to-keypoint scene encoding to direct encoder-to-grid pooling (ablation) | PV-RCNN · KITTI val, Car 3D Easy | mAP: 91.88 -> 92.57 | Helped |
| Add voxel-to-keypoint scene encoding to direct encoder-to-grid pooling (ablation) | PV-RCNN · KITTI val, Car 3D Moderate | mAP: 82.86 -> 84.83 | Helped |
| Add voxel-to-keypoint scene encoding to direct encoder-to-grid pooling (ablation) | PV-RCNN · KITTI val, Car 3D Hard | mAP: 80.52 -> 82.69 | Helped |
| Add multi-scale voxel features pv3 and pv4 to keypoint features (ablation) | PV-RCNN VSA · KITTI val, Car Moderate | mAP: 83.17 -> 84.54 | Helped |
| Add BEV features to pv3 and pv4 keypoint features (ablation) | PV-RCNN VSA · KITTI val, Car Moderate | mAP: 84.54 -> 84.69 | Helped |
| Add raw point features to pv3, pv4, and BEV keypoint features (ablation) | PV-RCNN VSA · KITTI val, Car Moderate | mAP: 84.69 -> 84.72 | No clear effect |
| Add pv1 and pv2 voxel features to the keypoint representation (ablation) | PV-RCNN VSA · KITTI val, Car Moderate | mAP: 84.75 -> 84.83 | Helped |
| Add predicted keypoint weighting (ablation) | PV-RCNN · KITTI val, Car 3D Moderate | mAP: 82.95 -> 84.83 | Helped |
| Replace RoI-aware pooling with RoI-grid pooling (ablation) | PV-RCNN · KITTI val, Car 3D Moderate | mAP: 82.97 -> 84.83 | Helped |
| Use IoU-guided confidence targets instead of classification targets (ablation) | PV-RCNN · KITTI val, Car 3D Moderate | mAP: 82.50 -> 84.83 | Helped |
Closest papers by meaning
- Point-Voxel CNN for Efficient 3D Deep Learning (NeurIPS 2019 · similarity 0.92)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.92)
- Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection (AAAI 2021 · similarity 0.91)
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.91)
- Voxel Set Transformer: A Set-to-Set Approach to 3D Object Detection From Point Clouds (CVPR 2022 · similarity 0.91)
- 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.