Research map / Point clouds / Segmentation and detection
Searching Efficient 3D Architectures with Sparse Point-Voxel Convolution
Haotian Tang, Zhijian Liu, Shengyu Zhao, Yujun Lin, Ji Lin, Hanrui Wang, Song Han
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 high-resolution point-based branch to Sparse Convolution (SPVCNN vs MinkowskiNet) (ablation) | MinkowskiNet · SemanticKITTI test (smallest scale) | mIoU: 57.5 -> 58.5 | Improved |
| add point-based branch (SPVCNN vs MinkowskiNet) (ablation) | MinkowskiNet · SemanticKITTI test (5.5M params) | mIoU: 60.0 -> 61.6 | Improved |
| add point-based branch (SPVCNN vs MinkowskiNet) (ablation) | MinkowskiNet · SemanticKITTI val seq 08 (21.7M params) | mIoU: 61.1 -> 63.8 | Improved |
| add point-based branch (SPVCNN vs MinkowskiNet) (ablation) | MinkowskiNet · SemanticKITTI test (21.7M params) | mIoU: 63.1 -> 63.8 | Improved |
| add 3D-NAS to SPVCNN (SPVNAS vs SPVCNN) (ablation) | SPVCNN · SemanticKITTI test (smallest scale) | mIoU: 58.5 -> 63.6 | Improved |
| use SPVNAS instead of MinkowskiNet | MinkowskiNet · SemanticKITTI test | mIoU: 63.1 -> 66.4 | Improved |
| add point-based branch to Sparse Convolution on small classes (ablation) | MinkowskiNet · SemanticKITTI test Motorcyclist | IoU: 18.7 -> 43.7 | Improved |
| add point-based branch to Sparse Convolution on small classes (ablation) | MinkowskiNet · SemanticKITTI test Bicyclist | IoU: 61.9 -> 65.2 | Improved |
| replace 3D sparse convolutions in SECOND with SPVConv (ablation) | SECOND (Repro.) · KITTI test Cyclist moderate | mAP: 59.7 -> 63.7 | Improved |
| replace 3D sparse convolutions in SECOND with SPVConv (ablation) | SECOND (Repro.) · KITTI test Pedestrian moderate | mAP: 41.7 -> 41.4 | No clear effect |
| replace 3D sparse convolutions in SECOND with SPVConv (ablation) | SECOND · KITTI val Pedestrian easy | mAP: 68.3 -> 68.2 | No clear effect |
| use evolutionary search instead of random search (ablation) | 3D-NAS · SemanticKITTI val | mIoU: around 3% better than first generation | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PolarNet — PolarNet: An Improved Grid Representation for Online LiDAR Point Clouds Semantic Segmentation (CVPR 2020) · better on 1 of 1 shared result
- RandLA-Net — RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds (CVPR 2020) · better on 1 of 1 shared result
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
- PVNAS: 3D Neural Architecture Search with Point-Voxel Convolution (TPAMI 2021 · similarity 0.92)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.90)
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.89)
- VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking (CVPR 2023 · similarity 0.89)
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.88)
- Point-Voxel CNN for Efficient 3D Deep Learning (NeurIPS 2019 · 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.
