Research map / Point clouds / Segmentation and detection
Point Density-Aware Voxels for LiDAR 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 |
|---|---|---|---|
| localize voxel features with point centroids instead of voxel centers (ablation) | PDV · Waymo validation, LEVEL_2 | mAPH: +0.83 vehicle; +3.86 pedestrian; +3.67 cyclist | Helped |
| add local feature density estimation using KDE (ablation) | PDV · Waymo validation, LEVEL_2 | mAPH: +0.07 vehicle; 4.45% pedestrian; +1.78% cyclist | Helped |
| add grid point self-attention (ablation) | PDV · Waymo validation, LEVEL_2 | mAPH: +0.45 vehicle; +0.61 pedestrian; +1.43 cyclist | Helped |
| add density confidence prediction (ablation) | PDV · Waymo validation, LEVEL_2 | mAPH: -0.01 vehicle; +0.23 pedestrian; +0.74 cyclist | Mixed |
| replace no positional encoding with sinusoidal encoding (ablation) | PDV density-aware RoI pooling · Waymo validation, LEVEL_2 | mAPH: -0.25 vehicle; +0.14 pedestrian; +0.50 cyclist | Mixed |
| use FFN spatial grid coordinates as positional encoding (ablation) | PDV density-aware RoI pooling · Waymo validation, LEVEL_2 | mAPH: +0.17 vehicle; +0.18 pedestrian; +0.27 cyclist | Helped |
| use point counts per grid voxel as positional encoding (ablation) | PDV density-aware RoI pooling · Waymo validation, LEVEL_2 | mAPH: +0.22 vehicle; +0.73 pedestrian; +0.67 cyclist | Helped |
| combine spatial coordinates and point counts in positional encoding (ablation) | PDV density-aware RoI pooling · Waymo validation, LEVEL_2 | mAPH: +0.15 vehicle; +1.07 pedestrian; +1.82 cyclist | Helped |
| use PDV rather than PV-RCNN | Waymo validation · LEVEL_2 mAPH averaged across classes | mAPH: +2.42% | Helped |
| use PDV rather than PV-RCNN | Waymo validation · inference | runtime: reducing runtime by 14% | Helped |
Closest papers by meaning
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.93)
- 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 0.91)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.90)
- JPV-Net: Joint Point-Voxel Representations for Accurate 3D Object Detection (AAAI 2022 · similarity 0.90)
- LiDAR-PTQ: Post-Training Quantization for Point Cloud 3D Object Detection (ICLR 2024 · similarity 0.89)
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.