Research map / Point clouds / Shape and surface completion
Deep Hybrid Self-Prior for Full 3D Mesh Generation
Xingkui Wei, Zhengqing Chen, Yanwei Fu, Zhaopeng Cui, Yinda 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 2D self-prior refinement on top of 3D prior (Point2Mesh as ablation without 2D prior) (ablation) | 3D self-prior mesh network · Synthetic 14 textured meshes | F-score: 97.2 -> 97.7 | Improved |
| add 2D self-prior refinement (ablation) | Point2Mesh · Thingi10K (sampled point clouds) | F-score: 93.4 -> 96.1 | Improved |
| add 2D self-prior refinement (ablation) | Point2Mesh · COSEG (sampled point clouds) | F-score: 92.8 -> 96.2 | Improved |
| add 2D self-prior refinement (ablation) | Point2Mesh · ShapeNet (sampled point clouds) | F-score: 91.4 -> 95.7 | Improved |
| add 2D self-prior refinement (ablation) | Point2Mesh · Synthetic 14 textured meshes | Chamfer distance: 0.0601 -> 0.0526 | Improved |
| hybrid prior vs Point2Mesh under 2% coordinate noise (ablation) | Point2Mesh · Synthetic, input noise 2% | F-score: 70.1 -> 83.5 | Improved |
| hybrid prior vs Point2Mesh under 10% coordinate noise (ablation) | Point2Mesh · Synthetic, input noise 10% | F-score: 18.0 -> 37.6 | Improved |
| add 2D UV-map texture network instead of MeshCNN per-vertex color (ablation) | texture reconstruction · Texture quality, rendered views | NIQE: 20.39 -> 19.35 | Improved |
| add edge length loss to 3D prior network (ablation) | 3D prior network · Convergence speed | steps to similar mesh quality: 4000 steps -> 1000 steps | Improved |
| expand UV atlas valid region by a few pixels (ablation) | 2D prior refinement · Mesh boundary artifacts | artifact removal | No clear effect |
| larger Gaussian perturbation in 2D prior input (ablation) | 2D prior network · Qualitative output | smoothness vs detail | Mixed |
| use distance UV map w.r.t. anchor point instead of XYZ map (ablation) | 2D prior network · Mesh refinement | geometry quality | Hurt |
Closest papers by meaning
- Leveraging 2D Data to Learn Textured 3D Mesh Generation (CVPR 2020 · similarity 0.90)
- Mesh2Tex: Generating Mesh Textures from Image Queries (ICCV 2023 · similarity 0.89)
- Convolutional Generation of Textured 3D Meshes (NeurIPS 2020 · similarity 0.88)
- Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors (ICLR 2024 · similarity 0.87)
- Self-Supervised Dual Contouring (CVPR 2024 · similarity 0.87)
- MeshFormer : High-Quality Mesh Generation with 3D-Guided Reconstruction Model (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.
