Research map / 3D vision / Face and shape
3DIAS: 3D Shape Reconstruction With Implicit Algebraic Surfaces
Mohsen Yavartanoo, Jaeyoung Chung, Reyhaneh Neshatavar, Kyoung Mu Lee
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 center location constraint (ablation) | 3DIAS · ShapeNet mean | F-Score: 46.72 -> 52.22 | Improved |
| add center location constraint (ablation) | 3DIAS · ShapeNet mean | Chamfer: 0.387 -> 0.197 | Improved |
| add scale (upper bound) constraint (ablation) | 3DIAS · ShapeNet mean | Chamfer: 0.261 -> 0.197 | Improved |
| add scale constraint (ablation) | 3DIAS · ShapeNet mean | IoU: 0.559 -> 0.575 | Improved |
| add closedness constraint on top of scale constraint (ablation) | 3DIAS · ShapeNet mean | F-Score: 44.44 -> 48.45 | Improved |
| add surface normal loss (ablation) | 3DIAS · ShapeNet mean | F-Score: 49.17 -> 52.22 | Improved |
| add surface sign loss (ablation) | 3DIAS · ShapeNet mean | F-Score: 43.77 -> 52.22 | Improved |
| add both surface losses (ablation) | 3DIAS · ShapeNet mean | IoU: 0.542 -> 0.575 | Improved |
| train on multiple classes instead of single class | 3DIAS · ShapeNet mean | F-Score: 50.60 -> 52.22 | Improved |
| train on multiple classes instead of single class | 3DIAS · ShapeNet mean | Chamfer: 0.493 -> 0.197 | Improved |
| train on multiple classes instead of single class | 3DIAS · ShapeNet mean | IoU: 0.575 -> 0.575 | No clear effect |
| single-class instead of multi-class training uses more primitives | 3DIAS · ShapeNet mean | number of primitives: 13.71 -> 21.07 | Mixed |
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.
- CvxNet — CvxNet: Learnable Convex Decomposition (CVPR 2020) · better on 3 of 3 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
- Learning to Infer Implicit Surfaces without 3D Supervision (NeurIPS 2019 · similarity 0.89)
- Neural Parts: Learning Expressive 3D Shape Abstractions With Invertible Neural Networks (CVPR 2021 · similarity 0.88)
- Neural Vector Fields: Implicit Representation by Explicit Learning (CVPR 2023 · similarity 0.88)
- NeAT: Learning Neural Implicit Surfaces With Arbitrary Topologies From Multi-View Images (CVPR 2023 · similarity 0.88)
- Learning Shape Primitives via Implicit Convexity Regularization (ICCV 2023 · similarity 0.88)
- ANIM: Accurate Neural Implicit Model for Human Reconstruction from a single RGB-D Image (CVPR 2024 · 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.
