Research map / Point clouds / Segmentation and detection
PV-RCNN++: Point-Voxel Feature Set Abstraction With Local Vector Representation for 3D Object Detection
Shaoshuai Shi, Li Jiang, Jiajun Deng, Zhe Wang, Chaoxu Guo, Jianping Shi, Xiaogang Wang, Hongsheng Li
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 VectorPool aggregation in VSA and RoI-grid pooling (replacing set abstraction) (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 65.92 -> 66.87 | Improved |
| add VectorPool aggregation (reduces memory) (ablation) | PV-RCNN++ · Waymo val | GPU memory (MB): 10617 -> 7583 | Improved |
| use voxel-to-keypoint VSA instead of UNet-decoder with RoI-grid pooling (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 66.05 -> 66.87 | Improved |
| use VSA instead of UNet-decoder when combined with RoI-aware pooling (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 65.77 -> 64.31 | Hurt |
| replace RoI-aware pooling with RoI-grid pooling (ablation) | PV-RCNN++ with VSA · Waymo val | mAPH L2 avg: 64.31 -> 66.87 | Improved |
| add predicted keypoint weighting (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 66.66 -> 66.87 | Improved |
| add proposal-centric keypoint filtering to FPS (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 65.87 -> 66.99 | Improved |
| add proposal-centric filtering (faster sampling) (ablation) | FPS keypoint sampling · Waymo val | sampling time: 133ms -> 27ms | Improved |
| use sectorized FPS instead of vanilla FPS after filtering (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 66.99 -> 66.87 | No clear effect |
| use sectorized FPS instead of random sampling (ablation) | PV-RCNN++ · Waymo val | mAPH L2 avg: 63.75 -> 66.87 | Improved |
| use interpolation instead of average pooling for sub-voxels (ablation) | VectorPool aggregation · Waymo val | mAPH L2 avg: 66.06 -> 66.87 | Improved |
| use interpolation instead of random selection for sub-voxels (ablation) | VectorPool aggregation · Waymo val | mAPH L2 avg: 66.29 -> 66.87 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PV-RCNN — PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020 · similarity 0.97)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.92)
- Point-Voxel CNN for Efficient 3D Deep Learning (NeurIPS 2019 · similarity 0.91)
- 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 0.90)
- 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.
