Research map / Point clouds / Segmentation and detection
PVT-SSD: Single-Stage 3D Object Detector With Point-Voxel Transformer
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 the point-voxel detector with lifting, feature alignment, and both token types (ablation) | PVT-SSD · Waymo validation subset, vehicle | APH: 62.37 -> 65.01 | Helped |
| Add voxel tokens (ablation) | PVT-SSD · Waymo validation subset, vehicle | APH: 62.98 -> 63.43 | Helped |
| Add point tokens (ablation) | PVT-SSD · Waymo validation subset, vehicle | APH: 63.43 -> 65.01 | Helped |
| Add feature alignment (ablation) | PVT-SSD query initialization · Waymo validation subset, vehicle | APH: +0.87 APH | Helped |
| Add lifting of sampled 2D voxels into 3D reference points (ablation) | PVT-SSD query initialization · Waymo validation subset, vehicle | APH: +0.61 APH | Helped |
| Use S-FPS instead of FPS, F-FPS, or Top-K (ablation) | PVT-SSD query initialization · Waymo validation subset, vehicle | APH: 65.01; 64.53; 53.76; 64.20 | Helped |
| Increase voxel-token radius from 6.4 to 8.0 (ablation) | PVT-SSD point-voxel Transformer · Waymo validation subset, vehicle | APH: 64.79 -> 65.01 | Helped |
| Increase point-token radius from 1.6 to 3.2 (ablation) | PVT-SSD point-voxel Transformer · Waymo validation subset, vehicle | APH: 64.21 -> 65.01 | Helped |
| Increase either token radius beyond 8.0 for voxels or 3.2 for points (ablation) | PVT-SSD point-voxel Transformer · Waymo validation subset, vehicle | APH: 65.01 -> 64.94; 65.01 -> 64.97 | Hurt |
| Use contextual relative positional encoding instead of no encoding, bias-relative, or absolute encoding (ablation) | PVT-SSD Transformer · Waymo validation subset, vehicle | APH: 63.29; 64.28; 63.76 -> 65.01 | Helped |
| Replace random neighbor sampling with sequential sampling (ablation) | Virtual Range Image ball query · Waymo validation subset, vehicle | APH: 65.01 -> 64.11 | Hurt |
| Use three input frames instead of one | PVT-SSD · Waymo validation, vehicle LEVEL_1 | 3D AP/APH: 79.16/78.72 -> 80.59/80.16 | Helped |
Closest papers by meaning
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.91)
- 3DSSD: Point-Based 3D Single Stage Object Detector (CVPR 2020 · similarity 0.90)
- Voxel Set Transformer: A Set-to-Set Approach to 3D Object Detection From Point Clouds (CVPR 2022 · similarity 0.90)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.90)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.88)
- Voxel Mamba: Group-Free State Space Models for Point Cloud based 3D Object Detection (NeurIPS 2024 · similarity 0.87)
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.