Research map / Point clouds / Point cloud classification
Stratified Transformer for 3D Point Cloud Segmentation
Check 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 stratified attention to the model with point embedding, augmentation, and cRPE (ablation) | Stratified Transformer · S3DIS | mIoU: 70.1 -> 72.0; +1.9% | Helped |
| Add stratified attention to the model with point embedding, augmentation, and cRPE (ablation) | Stratified Transformer · ScanNetv2 | mIoU: 72.5 -> 73.7; +1.2% | Helped |
| Add first-layer point embedding (ablation) | Stratified Transformer · S3DIS | mIoU: 70.0 -> 72.0; 2.0% mIoU gain | Helped |
| Add first-layer point embedding (ablation) | Stratified Transformer · ScanNetv2 | mIoU: 69.7 -> 73.7; 4.0% mIoU gain | Helped |
| Add cRPE (ablation) | Stratified Transformer · S3DIS | mIoU: 67.2 -> 70.1; 2.9% mIoU | Helped |
| Add cRPE (ablation) | Stratified Transformer · ScanNetv2 | mIoU: 70.6 -> 72.5; 1.9% mIoU | Helped |
| Replace cRPE with MLP-based relative position encoding (ablation) | Stratified Transformer · S3DIS | mIoU: 68.0 vs 68.0 | No clear effect |
| Add data augmentation (ablation) | Stratified Transformer · S3DIS | mIoU: 66.1 -> 72.0; +5.9 | Helped |
| Use KPConv rather than linear projection for first-layer local aggregation (ablation) | Stratified Transformer · S3DIS | mIoU: 68.9 -> 72.0; +3.1 | Helped |
| Apply shifted windows (ablation) | Stratified Transformer · S3DIS | mIoU: 70.1 -> 72.0 | Helped |
| Use stratified attention instead of vanilla window attention without shifts (ablation) | Transformer block · S3DIS | mIoU: 69.4 -> 70.1 | Helped |
| Apply cRPE to query, key, and value rather than MLP positional encoding (ablation) | Stratified Transformer · S3DIS | mIoU: 68.0 -> 72.0 | Helped |
Closest papers by meaning
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.89)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.88)
- Point-SAM: Promptable 3D Segmentation Model for Point Clouds (ICLR 2025 · similarity 0.88)
- OneFormer3D: One Transformer for Unified Point Cloud Segmentation (CVPR 2024 · similarity 0.88)
- X-3D: Explicit 3D Structure Modeling for Point Cloud Recognition (CVPR 2024 · similarity 0.88)
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (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.