Research map / Point clouds / Point cloud classification
Revisiting Point Cloud Shape Classification with a Simple and Effective Baseline
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 six orthogonal views instead of one view (ablation) | SimpleView · ModelNet40 validation | accuracy: 90.7 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.1 -> 92.9 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 | Helped |
| use six orthogonal views instead of three views (ablation) | SimpleView · ModelNet40 validation | accuracy: 92.1 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.2 -> 92.9 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 | Helped |
| use perspective instead of orthographic projection (ablation) | SimpleView · ModelNet40 validation | accuracy: 92.7 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 -> 92.9 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 | No clear effect |
| concatenate view features instead of pooling them (ablation) | SimpleView · ModelNet40 validation | accuracy: 91.8 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 -> 92.9 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 | Helped |
| use weighted-average depth instead of minimum depth (ablation) | SimpleView · ModelNet40 validation | accuracy: 92.8 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.4 -> 92.9 $\tiny \raisebox{.2ex}{$\scriptstyle\pm$}$ 0.3 | No clear effect |
| use smooth loss instead of cross-entropy (ablation) | PointNet, PointNet++, DGCNN, and RSCNN · ModelNet40 | accuracy | Helped |
| use random scaling and translation augmentation instead of the PointNet++ augmentation (ablation) | PointNet, PointNet++, DGCNN, and RSCNN · ModelNet40 | accuracy | Helped |
| select the best test-set model instead of using final model selection (ablation) | PointNet, PointNet++, DGCNN, and RSCNN · ModelNet40 | accuracy | Helped |
| switch from PointNet++ no Vote to RSCNN Vote protocol (ablation) | PointNet++ · ModelNet40 | accuracy: 89.8% -> 93.3% | Helped |
Closest papers by meaning
- Rethinking Point Cloud Data Augmentation: Topologically Consistent Deformation (ICML 2025 · similarity 0.88)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.88)
- SMART-PC: Skeletal Model Adaptation for Robust Test-Time Training in Point Clouds (ICML 2025 · similarity 0.88)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.88)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.88)
- An Explicit Frame Construction for Normalizing 3D Point Clouds (ICML 2024 · 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.