Research map / Point clouds / Segmentation and detection
JPV-Net: Joint Point-Voxel Representations for Accurate 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 |
|---|---|---|---|
| combine voxel and point representations with attentive feature fusion (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.07 -> 84.91 | Helped |
| replace attentive fusion with concatenation (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.91 -> 84.82 | Hurt |
| add point features and point-voxel fusion to the voxel-only network (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.48 -> 84.91 | Helped |
| add voxel-to-point feature projection and fusion to the point-only network (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.61 -> 84.91 | Helped |
| replace Trilinear SA with 3NN interpolation (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.91 -> 84.83 | Hurt |
| replace Trilinear SA with Voxel Set Abstraction (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.91 -> 84.94 | No clear effect |
| add BEV IoU-aware proposal generation (ablation) | JPV-Net proposal stage · KITTI val, Car | AP Moderate: 81.78 -> 81.92 | Helped |
| use BEV IoU rather than 3D IoU for proposal generation (ablation) | JPV-Net proposal stage · KITTI val, Car | AP Moderate: 81.87 -> 81.92 | No clear effect |
| add point-IoU-aware refinement (ablation) | JPV-Net refinement stage · KITTI val, Car | AP Moderate: 84.91 -> 85.29 | Helped |
| use point IoU rather than 3D IoU for refinement (ablation) | JPV-Net refinement stage · KITTI val, Car | AP Moderate: 85.15 -> 85.29 | Helped |
| add both IoU-aware modules (ablation) | JPV-Net · KITTI val, Car | AP Moderate: 84.91 -> 85.60 | Helped |
| use JPV-Net rather than PV-RCNN | KITTI test, Cyclist 3D · KITTI test | AP Moderate: 63.71 -> 65.41 | Helped |
Closest papers by meaning
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.93)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.92)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.91)
- HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection (CVPR 2020 · similarity 0.91)
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.90)
- Point Density-Aware Voxels for LiDAR 3D Object Detection (CVPR 2022 · similarity 0.90)
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.