Research map / Point clouds / Point cloud classification
Spherical Fractal Convolutional Neural Networks for Point Cloud Recognition
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 spherical fractal lattices and adaptive projection instead of direct point processing | SFCNN · ModelNet40 z/SO3 | accuracy: 85.3 vs 81.5 | Helped |
| Use aligned spherical coordinates in the projection module (ablation) | SFCNN · ModelNet40 z/SO3 | accuracy: 47.2 -> 83.2 | Helped |
| Add the projection invertibility constraint (ablation) | SFCNN · ModelNet40 z/z | accuracy: 90.2 -> 90.8 | Helped |
| Add the projection invertibility constraint (ablation) | SFCNN · ModelNet40 z/SO3 | accuracy: 83.2 -> 83.7 | Helped |
| Increase channel width from K = 4 to K = 8 (ablation) | SFCNN · ModelNet40 z/z | accuracy: 90.2 -> 90.8 | Helped |
| Increase channel width from K = 4 to K = 8 (ablation) | SFCNN · ModelNet40 z/SO3 | accuracy: 83.2 -> 84.7 | Helped |
| Increase block count from B = 2 to B = 3 (ablation) | SFCNN · ModelNet40 z/z | accuracy: 90.2 -> 90.7 | Helped |
| Increase block count from B = 2 to B = 3 (ablation) | SFCNN · ModelNet40 z/SO3 | accuracy: 83.2 -> 83.7 | Helped |
| Increase sampled neighbors per vertex from k = 16 to k = 64 (ablation) | SFCNN · ModelNet40 z/z | accuracy: 90.2 -> 89.5 | Hurt |
| Increase sampled neighbors per vertex from k = 16 to k = 64 (ablation) | SFCNN · ModelNet40 z/SO3 | accuracy: 83.2 -> 82.0 | Hurt |
| Reduce sampled neighbors per vertex from k = 16 to k = 4 (ablation) | SFCNN · ModelNet40 z/z | accuracy: 90.2 -> 89.7 | Hurt |
| Reduce sampled neighbors per vertex from k = 16 to k = 4 (ablation) | SFCNN · ModelNet40 z/SO3 | accuracy: 83.2 -> 83.5 | No clear effect |
Closest papers by meaning
- Point Cloud Pre-Training With Natural 3D Structures (CVPR 2022 · similarity 0.88)
- Rotation-Invariant Local-to-Global Representation Learning for 3D Point Cloud (NeurIPS 2020 · similarity 0.88)
- Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution (AAAI 2020 · similarity 0.88)
- Rotation Equivariant Graph Convolutional Network for Spherical Image Classification (CVPR 2020 · similarity 0.87)
- Deep Learning 3D Shapes Using Alt-az Anisotropic 2-Sphere Convolution (ICLR 2019 · similarity 0.87)
- TetraSphere: A Neural Descriptor for O(3)-Invariant Point Cloud Analysis (CVPR 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.