Research map / Detection and segmentation / 3D detection and driving
Spatial Pruned Sparse Convolution for Efficient 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 |
|---|---|---|---|
| replace sparse convs with SPS-Conv | CenterPoint · nuScenes val | GFLOPs: 62.9 -> 34.3 | Helped |
| replace sparse convs with SPS-Conv | CenterPoint · nuScenes val | mAP: 58.6 -> 58.5 | No clear effect |
| replace sparse convs with SPS-Conv | Voxel R-CNN · KITTI val Car | GFLOPs: 7.6 -> 3.6 | Helped |
| replace sparse convs with SPS-Conv | PV-RCNN · KITTI val Car Moderate | AP3D: 83.69 -> 83.22 | No clear effect |
| add SPSS + SPRS to CenterPoint | CenterPoint · Waymo val (20% data) | GFLOPs: 76.7 -> 28.8 | Helped |
| increase SPSS-Conv pruning ratio from 0.5 to 0.9 (ablation) | SPSS-Conv · nuScenes val | mAP: 58.19 -> 55.21 | Hurt |
| increase SPRS-Conv pruning ratio from 0.5 to 0.9 (ablation) | SPRS-Conv · nuScenes val | mAP: 58.58 -> 52.09 | Hurt |
| use magnitude-based selection instead of random drop (ablation) | SPSS-Conv · nuScenes val | mAP: 56.02 -> 58.48 | Helped |
| use magnitude-based selection instead of inverse selection (ablation) | SPRS-Conv · nuScenes val | mAP: 16.72 -> 58.58 | Helped |
| magnitude mask instead of learned conv3x3 mask (ablation) | SPSS-Conv · nuScenes val | mAP: 57.29 -> 58.19 | Helped |
| skip connection instead of average pooling for unselected positions (ablation) | SPSS-Conv · nuScenes val | mAP: 58.03 -> 58.19 | No clear effect |
Closest papers by meaning
- Text-guided Sparse Voxel Pruning for Efficient 3D Visual Grounding (CVPR 2025 · similarity 0.87)
- FSHNet: Fully Sparse Hybrid Network for 3D Object Detection (CVPR 2025 · similarity 0.87)
- OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmentation (CVPR 2024 · similarity 0.86)
- SP3D: Boosting Sparsely-Supervised 3D Object Detection via Accurate Cross-Modal Semantic Prompts (CVPR 2025 · similarity 0.86)
- SparseFormer: Sparse Visual Recognition via Limited Latent Tokens (ICLR 2024 · similarity 0.86)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.86)
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.