Research map / Point clouds / Point cloud classification
Spatial Transformer for 3D Point Clouds
Jiayun Wang, Rudrasis Chakraborty, Stella X. Yu
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 deformable spatial transformers | DGCNN edge convolution · ModelNet40 classification | accuracy: 88.8 -> 89.9 | Improved |
| add deformable spatial transformers with random input rotations | DGCNN · ModelNet40 (random rotations) | accuracy: 85.7 -> 88.3 | Improved |
| add deformable spatial transformers to SplatNet | SplatNet · ModelNet40 classification | accuracy: 86.3 -> 88.6 | Improved |
| add deformable spatial transformers | DGCNN · ShapeNet part segmentation | mIoU: 84.2 -> 85.3 | Improved |
| add projective transformers instead of affine (ablation) | DGCNN · ShapeNet part segmentation | mIoU: 84.7 -> 84.4 | Hurt |
| add deformable transformers to PointCNN | PointCNN · ShapeNet part segmentation | mIoU: 84.9 -> 85.8 | Improved |
| add deformable transformers | SplatNet · ShapeNet part segmentation | mIoU: 84.6 -> 85.2 | Improved |
| add deformable transformers | DGCNN · S3DIS | mIoU: 56.0 -> 57.2 | Improved |
| add deformable transformers | VoxelNet · KITTI car 3D Easy | AP: 45.7 -> 46.1 | Improved |
| update transformation matrices end-to-end instead of random fixed (ablation) | Deformable transformer · ShapeNet part segmentation | mIoU: 84.7 -> 85.3 | Improved |
| add more graphs per layer (1 to 4) at fixed sub-feature dim 32 (ablation) | Deformable transformer · ShapeNet part segmentation | mIoU: 84.9 -> 85.3 | Improved |
| add more graphs at fixed total dim 64 (1 to 4) (ablation) | Deformable transformer · ShapeNet part segmentation | mIoU: 85.3 -> 83.5 | Hurt |
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.
- VoxelNet — VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018) · better on 2 of 2 shared results
- PointCNN — PointCNN: Convolution On X-Transformed Points (NeurIPS 2018) · 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
- Cloud Transformers: A Universal Approach to Point Cloud Processing Tasks (ICCV 2021 · similarity 0.91)
- Fully-Convolutional Point Networks for Large-Scale Point Clouds (ECCV 2018 · similarity 0.90)
- LitePT: Lighter Yet Stronger Point Transformer (CVPR 2026 · similarity 0.90)
- A-CNN: Annularly Convolutional Neural Networks on Point Clouds (CVPR 2019 · similarity 0.90)
- Stratified Transformer for 3D Point Cloud Segmentation (CVPR 2022 · similarity 0.90)
- Point Transformer V3: Simpler Faster Stronger (CVPR 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.
