Research map / Point clouds / Point cloud classification
PatchFormer: An Efficient Point Transformer With Patch Attention
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 |
|---|---|---|---|
| Use 32 bases instead of 96 (ablation) | PatchFormer · ModelNet40 | OA: 93.52 -> 91.54 | Hurt |
| Use 64 bases instead of 96 (ablation) | PatchFormer · ModelNet40 | OA: 93.52 -> 92.94 | Hurt |
| Use 128 bases instead of 96 (ablation) | PatchFormer · ModelNet40 | OA: 93.52 -> 93.50 | No clear effect |
| Use 128 rather than 96 bases (ablation) | PatchFormer · ModelNet40 | latency: 34.32ms -> 35.56ms | Hurt |
| Remove multi-scale feature aggregation (ablation) | PatchFormer · ModelNet40 | OA: 93.52 -> 92.85 | Hurt |
| Remove multi-scale feature aggregation (ablation) | PatchFormer · ShapeNet | mIoU: 86.52 -> 85.22 | Hurt |
| Replace PAT with MLP (ablation) | PatchFormer · ModelNet40 | OA: 93.52 -> 92.62 | Hurt |
| Replace PAT with MLP (ablation) | PatchFormer · ShapeNet | mIoU: 86.52 -> 85.32 | Hurt |
| Replace PAT with self-attention (ablation) | PatchFormer · ModelNet40 | OA: 93.52 -> 93.29 | Hurt |
| Replace PAT with self-attention (ablation) | PatchFormer · ShapeNet | mIoU: 86.52 -> 86.22 | Hurt |
| Remove 3D relative position bias (ablation) | PatchFormer · ModelNet40 | OA: +0.37% OA | Hurt |
| Remove 3D relative position bias (ablation) | PatchFormer · ShapeNet | mIoU: +0.47% mIoU | Hurt |
Closest papers by meaning
- FlatFormer: Flattened Window Attention for Efficient Point Cloud Transformer (CVPR 2023 · similarity 0.90)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.87)
- Dual Focus-Attention Transformer for Robust Point Cloud Registration (CVPR 2025 · similarity 0.86)
- PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning (CVPR 2025 · similarity 0.85)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.85)
- PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priors (CVPR 2025 · similarity 0.85)
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.