Research map / 3D vision / Face and shape
Surface Networks via General Covers
Niv Haim, Nimrod Segol, Heli Ben-Hamu, Haggai Maron, Yaron Lipman
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 |
|---|---|---|---|
| use covering-map toric image representation instead of octahedron unfolding for spherical signals | Inception v3 on ModelNet40 (rot z) · ModelNet40 classification | accuracy (%): 90.2% -> 91.0% | Improved |
| use covering-map toric image representation instead of equirectangular projection for spherical signals | Inception v3 on ModelNet40 (rot z) · ModelNet40 classification | accuracy (%): 90.1% -> 91.0% | Improved |
| train and test in original orientation instead of random z rotation | covering-map method · ModelNet40 classification | accuracy (%): 91.0% -> 91.6% | Improved |
| use covering-map representation instead of Spherical CNN | SHREC17 retrieval · SHREC17 retrieval (perturbed, randomly rotated) | F1@N: 0.699 -> 0.734 | Improved |
| use covering-map representation instead of SO(3) equivariant CNNs | SHREC17 retrieval · SHREC17 retrieval (perturbed, randomly rotated) | mAP: 0.685 -> 0.709 | Improved |
| use covering-map method instead of best retrieval entry FURUYA_DLAN | SHREC17 retrieval · SHREC17 retrieval (perturbed, randomly rotated) | P@N: 0.814 -> 0.749 | Hurt |
| apply covering-map U-Net on 3D coordinates instead of Pointnet++ | human segmentation · Human body segmentation | accuracy (%): 90.77% -> 91.31% | Improved |
| use covering-map method instead of Dynamic graph CNN on segmentation | human segmentation · Human body segmentation | accuracy (%): 89.72% -> 91.31% | Improved |
| use covering-map method instead of PCNN on ModelNet40 | ModelNet40 classification · ModelNet40 classification | accuracy (%): 92.3% -> 91.6% | Hurt |
Closest papers by meaning
- Deep Learning 3D Shapes Using Alt-az Anisotropic 2-Sphere Convolution (ICLR 2019 · similarity 0.89)
- Neural Convolutional Surfaces (CVPR 2022 · similarity 0.88)
- SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks (CVPR 2017 · similarity 0.87)
- PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations (ECCV 2020 · similarity 0.87)
- Field Convolutions for Surface CNNs (ICCV 2021 · similarity 0.86)
- Improving Neural Network Surface Processing with Principal Curvatures (NeurIPS 2024 · 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.
