Research map / 3D vision / Face and shape
SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use geometry images with shared correspondence instead of independently parameterized shapes (ablation) | SurfNet rigid-shape generation · Shape-category geometry image generation | geometry image coherence | Helped |
| Predict x, y, and z channels with separate networks instead of one network (ablation) | SurfNet geometry image generation · Geometry image generation | training error | Helped |
| Add curvature-based weights to the reconstruction loss (ablation) | SurfNet · Rigid and non-rigid shape reconstruction | edge preservation | Helped |
| Predict residual geometry images and add them to a base shape (ablation) | SurfNet rigid-shape generation · Rigid shapes from class and view parameters | reconstruction error | Helped |
| Increase geometry-image resolution from 64 × 64 to 128 × 128 (ablation) | SurfNet · Geometry image generation | training time | Hurt |
| Reconstruct surfaces directly from depth images | SurfNet · Hand depth-image test set | surface reconstruction quality | Helped |
| Reconstruct surfaces directly from RGB images with implicit viewpoint estimation | SurfNet · Car and airplane RGB-image test sets | surface reconstruction quality | Helped |
| Use SurfNet rather than Kar et al.'s method | PASCAL 3D+ car reconstruction · PASCAL 3D+ car | surface reconstruction quality | Helped |
| Use SurfNet rather than Kar et al.'s method | PASCAL 3D+ aeroplane reconstruction · PASCAL 3D+ aeroplane | surface reconstruction quality | Helped |
| Interpolate between parametric hand poses | SurfNet hand generator · Hand pose interpolation | surface transition quality | Helped |
| Interpolate between shape codes while varying view angle | SurfNet rigid-shape generator · Car and airplane shape interpolation | intermediate surface quality | Helped |
| Use image inputs with low contrast, poor texture, or viewpoints outside training ranges | SurfNet image reconstruction · Unseen rigid-shape images | surface reconstruction quality | Hurt |
Closest papers by meaning
- A Papier-Mâché Approach to Learning 3D Surface Generation (CVPR 2018 · similarity 0.91)
- Shape Completion Using 3D-Encoder-Predictor CNNs and Shape Synthesis (CVPR 2017 · similarity 0.90)
- Synthesizing 3D Shapes via Modeling Multi-View Depth Maps and Silhouettes With Deep Generative Networks (CVPR 2017 · similarity 0.89)
- Geometry-Aware Network for Non-Rigid Shape Prediction From a Single View (CVPR 2018 · similarity 0.89)
- Neural Convolutional Surfaces (CVPR 2022 · similarity 0.89)
- Learning to Infer Implicit Surfaces without 3D Supervision (NeurIPS 2019 · similarity 0.89)
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.