Research map / Point clouds / Segmentation and detection
Spherical Transformer for LiDAR-Based 3D Recognition
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 local sparse-convolution aggregation with radial window attention | SparseConv semantic segmentation model · Waymo Open Dataset val | mIoU: 66.6 -> 69.9 | Helped |
| Add radial window attention (ablation) | SparseConv on nuScenes val · nuScenes val, far points | mIoU: 13.28 -> 26.67 | Helped |
| Use radial rather than cubic windows (ablation) | SphereFormer on nuScenes val · nuScenes val | overall mIoU: 76.19 -> 78.41 | Helped |
| Add exponential splitting to radial attention (ablation) | SphereFormer on nuScenes val · nuScenes val | overall mIoU: 77.05 -> 78.41 | Helped |
| Add exponential splitting to radial attention (ablation) | SphereFormer on nuScenes val · nuScenes val, far points | mIoU: 26.67 -> 31.10 | Helped |
| Add dynamic feature selection to radial attention and exponential splitting (ablation) | SphereFormer on nuScenes val · nuScenes val | overall mIoU: 77.60 -> 78.41 | Helped |
| Add dynamic feature selection to radial attention (ablation) | SphereFormer on nuScenes val · nuScenes val | overall mIoU: 76.31 -> 77.05 | Helped |
| Add dynamic feature selection to radial attention (ablation) | SphereFormer on nuScenes val · nuScenes val, medium and far points | mIoU: 57.21 -> 58.94; 26.67 -> 28.95 | Mixed |
| Set the angular window size to 1.0° rather than 2.0° (ablation) | SphereFormer on nuScenes val · nuScenes val | mIoU: 78.4 -> 77.8 | Hurt |
| Set the angular window size to 2.5° rather than 2.0° (ablation) | SphereFormer on nuScenes val · nuScenes val | mIoU: 78.4 -> 77.6 | Hurt |
| Add rotation and translation test-time augmentation | SphereFormer semantic segmentation · nuScenes val | mIoU: 78.4 -> 79.5 | Helped |
| Add flipping and rotation test-time augmentation | SphereFormer object detection · nuScenes test | NDS: 70.7 -> 72.8 | Helped |
Closest papers by meaning
- Spherical Frustum Sparse Convolution Network for LiDAR Point Cloud Semantic Segmentation (NeurIPS 2024 · similarity 0.88)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.87)
- TULIP: Transformer for Upsampling of LiDAR Point Clouds (CVPR 2024 · similarity 0.87)
- LSK3DNet: Towards Effective and Efficient 3D Perception with Large Sparse Kernels (CVPR 2024 · similarity 0.87)
- HOTFormerLoc: Hierarchical Octree Transformer for Versatile Lidar Place Recognition Across Ground and Aerial Views (CVPR 2025 · similarity 0.87)
- GBlobs: Explicit Local Structure via Gaussian Blobs for Improved Cross-Domain LiDAR-based 3D Object Detection (CVPR 2025 · 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.