Research map / Point clouds / Point cloud classification
GPSFormer: A Global Perception and Local Structure Fitting-based Transformer for Point Cloud Understanding
Changshuo Wang, Meiqing Wu, Siew-Kei Lam, Xin Ning, Shangshu Yu, Ruiping Wang, Weijun Li, Thambipillai Srikanthan
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 |
|---|---|---|---|
| combine ADGConv, RCA and MHA in the Global Perception Module (ablation) | GPSFormer GPM · ScanObjectNN PB_T50_RS | overall accuracy: 93.2 (ADGConv only) -> 95.4 | Improved |
| add RCA to ADGConv (ablation) | GPSFormer GPM · ScanObjectNN PB_T50_RS | overall accuracy: 93.2 -> 94.4 | Improved |
| add MHA to ADGConv (ablation) | GPSFormer GPM · ScanObjectNN PB_T50_RS | overall accuracy: 93.2 -> 95.0 | Improved |
| vary ADGConv neighborhood size, 20 versus 5 or 30 (ablation) | ADGConv · ScanObjectNN PB_T50_RS | overall accuracy: 88.6 (k=5), 95.4 (k=20), 93.2 (k=30) | Mixed |
| use learnable polynomial order terms (s=1, p learnable) instead of ABF or RBF (ablation) | LSFConv · ScanObjectNN PB_T50_RS | accuracy: 92.8 (ABF) / 93.2 (RBF) -> 95.4 | Improved |
| train from scratch rather than pre-trained on few-shot ScanObjectNN | GPSFormer · ScanObjectNN few-shot 5-way 10-shot | accuracy: 89.3 -> 71.7 | Hurt |
| apply test-time voting (ablation) | GPSFormer · ScanObjectNN PB_T50_RS | overall accuracy: 95.0 -> 95.4 | Improved |
| use GPSFormer instead of larger supervised models | point cloud classifier · ModelNet40 | overall accuracy: 94.2 vs PointGPT 94.9 and RepSurf-U 94.7 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports GPSFormer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Point Clouds Meets Physics: Dynamic Acoustic Field Fitting Network for Point Cloud Understanding (CVPR 2025 · 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
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.89)
- Spatial Transformer for 3D Point Clouds (TPAMI 2021 · similarity 0.89)
- ConDaFormer: Disassembled Transformer with Local Structure Enhancement for 3D Point Cloud Understanding (NeurIPS 2023 · similarity 0.89)
- Multi-Scale Part-Based Feature Representation for 3D Domain Generalization and Adaptation (TPAMI 2024 · similarity 0.89)
- A Linear Time and Space Local Point Cloud Geometry Encoder via Vectorized Kernel Mixture (VecKM) (ICML 2024 · similarity 0.88)
- PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priors (CVPR 2025 · similarity 0.88)
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.
