Research map / Point clouds / Point cloud classification
Self-Positioning Point-Based Transformer for Point Cloud Understanding
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 |
|---|---|---|---|
| use SPoTr instead of the strongest listed baseline | Shape classification on SONN_PB_T50_RS · SONN_PB_T50_RS | mAcc: 85.8±0.6 -> 86.8 | Helped |
| use SPoTr instead of the strongest listed baseline | Shape classification on SONN_PB_T50_RS · SONN_PB_T50_RS | OA: 87.7±0.4 -> 88.6 | Helped |
| use SPoTr instead of PointNeXt | Part segmentation · SN-Part | ins. mIoU: 87.0 $\pm$ 0.1 -> 87.2 | Helped |
| use SPoTr instead of PointNeXt | Semantic segmentation · S3DIS Area-5 | mIoU: 70.5 ± 0.3 -> 70.8 | Helped |
| add SPA to local point attention (ablation) | SPoTr on SONN_PB · SONN_PB | OA: 87.9 -> 88.6 | Helped |
| use adaptive self-positioning points instead of FPS-selected points (ablation) | SPoTr on SONN_PB · SONN_PB | OA: 87.7 -> 88.6 | Helped |
| add disentangled spatial and semantic attention instead of using only the spatial kernel (ablation) | SPoTr on SONN_PB · SONN_PB | OA: 88.2 -> 88.6 | Helped |
| use CWPA instead of standard attention (ablation) | SPoTr on SONN_PB · SONN_PB | OA: 86.1 -> 88.6 | Helped |
| use f_k as the semantic relation instead of f_q - f_k (ablation) | CWPA in SPoTr · SONN_PB | OA: 88.6 -> 88.1 | Hurt |
| use f_q + f_k as the semantic relation instead of f_q - f_k (ablation) | CWPA in SPoTr · SONN_PB | OA: 88.6 -> 86.4 | Hurt |
| use f_q ⊙ f_k as the semantic relation instead of f_q - f_k (ablation) | CWPA in SPoTr · SONN_PB | OA: 88.6 -> 85.4 | Hurt |
Closest papers by meaning
- Modify Self-Attention via Skeleton Decomposition for Effective Point Cloud Transformer (AAAI 2022 · similarity 0.89)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.87)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.86)
- Dual Focus-Attention Transformer for Robust Point Cloud Registration (CVPR 2025 · similarity 0.86)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.86)
- Scene Flow Prior Based Point Cloud Completion with Masked Transformer (Student Abstract) (AAAI 2024 · 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.