Research map / 3D vision / Face and shape
Surface Snapping Optimization Layer for Single Image Object Shape Reconstruction
Yuan-Ting Hu, Alex Schwing, Raymond A. Yeh
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 surface snapping layer after each refinement module | Mesh R-CNN · Pix3D S1 and S2 (Setup 1) | APmesh: +0.8 (S1), +1.6 (S2) | Improved |
| add surface snapping layer after each refinement module | GCN Transformer · Pix3D S1 (Setup 1) | APmesh: -0.4 | Hurt |
| add surface snapping layer after each refinement module | GCN Transformer · Pix3D S2 (Setup 1) | APmesh: +0.5 | Improved |
| add surface snapping layer after each refinement module | Pixel2Mesh+ · Pix3D S2 (Setup 1) | APmesh: +1.0 | Improved |
| add surface snapping layer after marching cubes | IM3D · Pix3D (Setup 2) | Chamfer distance avg (x10^3): -0.08 (7.78 to 7.70) | Improved |
| add surface snapping layer | Mesh R-CNN · ShapeNet Best | Normal consistency: +0.002 | Improved |
| add surface snapping layer | Mesh R-CNN · ShapeNet Best | F1@0.1: -1.2 | Hurt |
| add surface snapping layer | Mesh R-CNN · ShapeNet Pretty | Chamfer distance: -0.005 (0.176 to 0.171) | Improved |
| learn the snapping strength alpha vs fixed alpha=0.0 (ablation) | Snap+ Mesh R-CNN · Pix3D S1 | APmesh: +0.7 (53.4 to 54.1) | Improved |
| fixed alpha=2.0 instead of learned alpha (ablation) | Snap+ Mesh R-CNN · Pix3D S1 | APmesh: -1.4 vs learned alpha | Hurt |
| learn alpha vs fixed alpha=0.0 (ablation) | Snap+ Mesh R-CNN · Pix3D S1 | Normal consistency: +1.5 (21.5 to 23.0) | Improved |
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 1 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
- Front2Back: Single View 3D Shape Reconstruction via Front to Back Prediction (CVPR 2020 · similarity 0.88)
- Image-Based 3D Object Reconstruction: State-of-the-Art and Trends in the Deep Learning Era (TPAMI 2019 · similarity 0.87)
- ARO-Net: Learning Implicit Fields From Anchored Radial Observations (CVPR 2023 · similarity 0.87)
- Pix2Surf: Learning Parametric 3D Surface Models of Objects from Images (ECCV 2020 · similarity 0.87)
- Deep Optimized Priors for 3D Shape Modeling and Reconstruction (CVPR 2021 · similarity 0.87)
- PMNI: Pose-free Multi-view Normal Integration for Reflective and Textureless Surface Reconstruction (CVPR 2025 · 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.
