Research map / Point clouds / Point cloud classification
Relation-Shape Convolutional Neural Network for Point Cloud Analysis
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 geometric relation learning (ablation) | RS-CNN · ModelNet40 | accuracy (%): 87.2 -> 89.9 | Helped |
| add batch normalization (ablation) | RS-CNN with relation learning · ModelNet40 | accuracy (%): 89.9 -> 91.9 | Helped |
| add dropout in fully connected layers (ablation) | RS-CNN with batch normalization · ModelNet40 | accuracy (%): 91.9 -> 92.2 | Helped |
| increase relation learning from one scale to two scales (ablation) | RS-CNN · ModelNet40 | accuracy (%): 92.2 -> 92.5 | Helped |
| increase relation learning from one scale to three scales (ablation) | RS-CNN · ModelNet40 | accuracy (%): 92.2 -> 92.9 | Helped |
| add ten-vote testing (ablation) | three-scale RS-CNN · ModelNet40 | accuracy (%): 92.9 -> 93.6 | Helped |
| increase input point count from 1k to 2k (ablation) | RS-CNN model G · ModelNet40 | accuracy (%): 93.6 -> 93.6 | No clear effect |
| remove geometric relation learning (ablation) | RS-CNN with the other techniques retained · ModelNet40 | accuracy (%): 90.1 -> 93.6 | Hurt |
| replace max pooling with average pooling (ablation) | RS-CNN · ModelNet40 | accuracy (%): 93.6 -> 91.6 | Hurt |
| replace max pooling with summation (ablation) | RS-CNN · ModelNet40 | accuracy (%): 93.6 -> 91.6 | Hurt |
| increase mapping MLP depth from three layers to four (ablation) | RS-CNN · ModelNet40 | accuracy (%): 93.6 -> 92.7 | Hurt |
| use only 3D Euclidean distance instead of the full 10-channel relation (ablation) | RS-CNN · ModelNet40 | accuracy (%): 92.5 -> 93.6 | Hurt |
Closest papers by meaning
- Shape-Oriented Convolution Neural Network for Point Cloud Analysis (AAAI 2020 · similarity 0.92)
- A-CNN: Annularly Convolutional Neural Networks on Point Clouds (CVPR 2019 · similarity 0.89)
- Attentional ShapeContextNet for Point Cloud Recognition (CVPR 2018 · similarity 0.88)
- Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling (CVPR 2018 · similarity 0.88)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.86)
- Rethinking Point Cloud Data Augmentation: Topologically Consistent Deformation (ICML 2025 · similarity 0.86)
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.