Research map / 3D vision / Face and shape
Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images
Nanyang Wang, Yinda Zhang, Zhuwen Li, Yanwei Fu, Wei Liu, Yu-Gang Jiang
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 ResNet-style features (ablation) | Pixel2Mesh · ShapeNet (Hausdorff ablation) | Hausdorff distance: 0.644 -> 0.603 | Improved |
| add Laplacian regularization (ablation) | Pixel2Mesh · ShapeNet (Hausdorff ablation) | Hausdorff distance: 0.600 -> 0.603 | Hurt |
| add graph unpooling (ablation) | Pixel2Mesh · ShapeNet (Hausdorff ablation) | Hausdorff distance: 0.616 -> 0.603 | Improved |
| add normal loss (ablation) | Pixel2Mesh · ShapeNet (Hausdorff ablation) | Hausdorff distance: 0.623 -> 0.603 | Improved |
| add edge length loss (ablation) | Pixel2Mesh · ShapeNet (Hausdorff ablation) | Hausdorff distance: 0.643 -> 0.603 | Improved |
| use horizontal ellipsoid instead of sphere as initial mesh (ablation) | Pixel2Mesh · ShapeNet (initial mesh) | F-score(tau): 56.93 -> 59.72 | Improved |
| use horizontal ellipsoid instead of sphere as initial mesh (ablation) | Pixel2Mesh · ShapeNet (initial mesh) | CD: 0.651 -> 0.591 | Improved |
| use horizontal ellipsoid instead of vertical ellipsoid as initial mesh (ablation) | Pixel2Mesh · ShapeNet (initial mesh) | F-score(tau): 56.58 -> 59.72 | Improved |
| use noisy ellipsoid instead of horizontal ellipsoid as initial mesh (ablation) | Pixel2Mesh · ShapeNet (initial mesh) | F-score(tau): 59.72 -> 59.25 | No clear effect |
| use Pixel2Mesh instead of octree voxel method (128^3) | Tartarchenko et al · ShapeNet-cars | F-score(tau): 65.335 -> 72.128 | Improved |
| use Pixel2Mesh instead of octree voxel method (128^3) | Tartarchenko et al · ShapeNet-cars | CD: 0.361 -> 0.236 | Improved |
| use Pixel2Mesh instead of octree voxel method (128^3) | Tartarchenko et al · ShapeNet-cars | EMD: 1.273 -> 1.220 | Improved |
Later papers that compared with this method
21 later accepted papers report Pixel2Mesh as a baseline in their results tables; 18 of them report a better number on most of the results they share with it.
- DeepMesh: Differentiable Iso-Surface Extraction (TPAMI 2024 · better on 4 of 4 shared results)
- Surface Snapping Optimization Layer for Single Image Object Shape Reconstruction (ICML 2023 · better on 1 of 1 shared result)
- Representing 3D Shapes With Probabilistic Directed Distance Fields (CVPR 2022 · better on 0 of 6 shared results)
- Structure-Preserving 3D Garment Modeling with Neural Sewing Machines (NeurIPS 2022 · better on 3 of 3 shared results)
- SAIL-VOS 3D: A Synthetic Dataset and Baselines for Object Detection and 3D Mesh Reconstruction From Video Data (CVPR 2021 · better on 6 of 8 shared results)
- Deep Active Surface Models (CVPR 2021 · better on 5 of 5 shared results)
- Learning Manifold Patch-Based Representations of Man-Made Shapes (ICLR 2021 · better on 2 of 4 shared results)
- Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis (NeurIPS 2021 · better on 1 of 2 shared results)
13 more not listed.
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
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction (NeurIPS 2025 · similarity 0.85)
- PivotMesh: Generic 3D Mesh Generation via Pivot Vertices Guidance (ICLR 2025 · similarity 0.84)
- One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion (CVPR 2024 · similarity 0.84)
- Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors (ICLR 2024 · similarity 0.84)
- MeshAnything: Artist-Created Mesh Generation with Autoregressive Transformers (ICLR 2025 · similarity 0.83)
- High-fidelity 3D Object Generation from Single Image with RGBN-Volume Gaussian Reconstruction Model (CVPR 2025 · similarity 0.83)
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.
