Research map / 3D vision / Face and shape
Deep Mesh Reconstruction From Single RGB Images via Topology Modification Networks
Junyi Pan, Xiaoguang Han, Weikai Chen, Jiapeng Tang, Kui Jia
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 topology modification (face pruning) and boundary refinement to the baseline (ablation) | Baseline mesh deformation · ShapeNet 5 categories mean | CD (x1e-3): 2.930 -> 2.277 | Improved |
| add topology modification and boundary refinement to the baseline (ablation) | Baseline mesh deformation · ShapeNet 5 categories mean | EMD (x1e-2): 10.360 -> 9.059 | Improved |
| add topology modification and boundary refinement (ablation) | Baseline · ShapeNet chair | CD (x1e-3): 4.325 -> 3.064 | Improved |
| add topology modification and boundary refinement (ablation) | Baseline · ShapeNet table | CD (x1e-3): 4.798 -> 3.616 | Improved |
| add topology modification and boundary refinement (ablation) | Baseline · ShapeNet plane | CD (x1e-3): 1.433 -> 1.390 | No clear effect |
| add topology modification and boundary refinement (ablation) | Baseline · ShapeNet firearm | CD (x1e-3): 1.145 -> 1.142 | No clear effect |
| use the proposed method instead of AtlasNet-25 before PSR on single-view reconstruction | AtlasNet-25 · ShapeNet mean | CD (x1e-3): 2.747 -> 2.277 | Improved |
| apply Poisson surface reconstruction to closed meshes, proposed vs AtlasNet-25 | AtlasNet-25 (PSR) · ShapeNet chair (20 shapes) | CD: 3.430 -> 2.304 | Improved |
| use the proposed method instead of AtlasNet-25 for autoencoding | AtlasNet-25 (AE) · ShapeNet autoencoding | CD: 0.765 -> 0.655 | Improved |
| use the proposed method instead of AtlasNet-25 for autoencoding | AtlasNet-25 (AE) · ShapeNet autoencoding | EMD: 8.467 -> 6.754 | Improved |
| add progressive shaping (keep the first subnet) before face pruning (ablation) | Ours · ShapeNet | qualitative mesh quality | Improved |
| use unit square instead of sphere as initial mesh (ablation) | Ours · ShapeNet | qualitative mesh quality | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Pixel2Mesh — Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images (ECCV 2018) · better on 2 of 2 shared results
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
- Neural Template: Topology-Aware Reconstruction and Disentangled Generation of 3D Meshes (CVPR 2022 · similarity 0.90)
- One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization (NeurIPS 2023 · similarity 0.89)
- SkeletonNet: A Topology-Preserving Solution for Learning Mesh Reconstruction of Object Surfaces From RGB Images (TPAMI 2021 · similarity 0.88)
- Pixel2Mesh: 3D Mesh Model Generation via Image Guided Deformation (TPAMI 2020 · similarity 0.88)
- Diffeomorphic Mesh Deformation via Efficient Optimal Transport for Cortical Surface Reconstruction (ICLR 2024 · similarity 0.86)
- QUADify: Extracting Meshes with Pixel-level Details and Materials from Images (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.
