Research map / 3D vision / Face and shape
DualConvMesh-Net: Joint Geodesic and Euclidean Convolutions on 3D Meshes
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 |
|---|---|---|---|
| combine geodesic and Euclidean radius neighborhoods in parallel (ablation) | DCM-Net with QEM pooling · ScanNet validation | mIoU: +3.4 | Helped |
| combine geodesic and Euclidean k-nn neighborhoods in parallel (ablation) | DCM-Net with QEM pooling · ScanNet validation | mIoU: +2.0 | Helped |
| add geodesic convolution to Euclidean radius convolution (ablation) | DCM-Net with Vertex Clustering · ScanNet validation | mIoU: +0.9 | Helped |
| add geodesic convolution to Euclidean k-nn convolution (ablation) | DCM-Net with Vertex Clustering · ScanNet validation | mIoU: -0.4 | Hurt |
| replace QEM pooling with Vertex Clustering (ablation) | DCM-Net using radius/geodesic neighborhoods · ScanNet validation | mIoU: +4.5 | Hurt |
| replace QEM pooling with Farthest Point Sampling (ablation) | SingleConvMesh-Net using radius neighborhoods · ScanNet validation | mIoU: +0.4 | Hurt |
| increase the test-time neighborhood threshold from T=15 to T=35 (ablation) | the segmentation model · S3DIS Area-5 | mIoU: +1.1% | Helped |
| increase the mesh hierarchy from 2 to 4 levels (ablation) | DCM-Net with QEM and radius/geodesic neighborhoods · ScanNet validation | mIoU: +12.9 | Helped |
| replace ReLU with Leaky ReLU (ablation) | DCM-Net with QEM and radius/geodesic neighborhoods · ScanNet validation | mIoU: +1.4 | Hurt |
| use 75% geodesic filters early and 25% geodesic filters in later levels (ablation) | DCM-Net · ScanNet validation | mIoU: 68.3 (± 0.12) | Helped |
Closest papers by meaning
- Primal-Dual Mesh Convolutional Neural Networks (NeurIPS 2020 · similarity 0.91)
- Modeling Local Geometric Structure of 3D Point Clouds Using Geo-CNN (CVPR 2019 · similarity 0.87)
- TextureNet: Consistent Local Parametrizations for Learning From High-Resolution Signals on Meshes (CVPR 2019 · similarity 0.87)
- PFCNN: Convolutional Neural Networks on 3D Surfaces Using Parallel Frames (CVPR 2020 · similarity 0.87)
- Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs (CVPR 2017 · similarity 0.87)
- DeCoTR: Enhancing Depth Completion with 2D and 3D Attentions (CVPR 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.