Research map / 3D vision / Face and shape
Laplacian Mesh Transformer: Dual Attention and Topology Aware Network for 3D Mesh Classification and Segmentation
Xiao-Juan Li, Jie Yang, Fang-Lue Zhang
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 |
|---|---|---|---|
| add the dual attention mechanism (two extractors plus fusion) (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 92.2 -> 97.7 | Improved |
| add the attention-based fusion module instead of simple concatenation (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 93.6 -> 97.7 | Improved |
| add Laplacian eigenvectors as topology input instead of coordinates (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 93.1 -> 97.7 | Improved |
| use 12 eigenvectors instead of 6 (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 93.8 -> 97.7 | Improved |
| use 12 eigenvectors instead of 18 (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 96.4 -> 97.7 | Improved |
| use 4 self-attention blocks instead of 2 (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 95.4 -> 97.7 | Improved |
| use 4 self-attention blocks instead of 3 (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 96.9 -> 97.7 | Improved |
| use 4 self-attention blocks instead of 5 (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 96.7 -> 97.7 | Improved |
| change mesh vertex count to 2000, 3500 or 5000 (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 97.7 -> 97.5, 97.2, 97.4 | No clear effect |
| re-mesh the data to change triangulation (ablation) | Laplacian Mesh Transformer · COSEG (mean) | segmentation accuracy: 97.7 -> 97.6 | No clear effect |
| use Laplacian Mesh Transformer instead of LaplacianNet for ShapeNet segmentation | ShapeNet part segmentation · ShapeNet | mean accuracy: 91.5 -> 92.6 | Improved |
| use Laplacian Mesh Transformer instead of LaplacianNet on ShapeNet classification | ShapeNet classification · ShapeNet classification | overall accuracy: 99.8 -> 99.4 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SyncSpecCNN — SyncSpecCNN: Synchronized Spectral CNN for 3D Shape Segmentation (CVPR 2017) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SchellingFormer: Laplacian Matrix-guided Geometric Transformer for Robust Schelling Point Detection (AAAI 2026 · similarity 0.88)
- Heat Diffusion Based Multi-Scale and Geometric Structure-Aware Transformer for Mesh Segmentation (CVPR 2023 · similarity 0.88)
- Primal-Dual Mesh Convolutional Neural Networks (NeurIPS 2020 · similarity 0.88)
- Spectral Shape Recovery and Analysis Via Data-driven Connections (IJCV 2021 · similarity 0.88)
- Learning Local Neighboring Structure for Robust 3D Shape Representation (AAAI 2021 · similarity 0.87)
- DMNet: Delaunay Meshing Network for 3D Shape Representation (ICCV 2023 · similarity 0.87)
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.
