Research map / 3D vision / Face and shape
Building 3D Generative Models from Minimal Data
Skylar Sutherland, Bernhard Egger, Joshua B. Tenenbaum
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 |
|---|---|---|---|
| combine RGB-space and physical-space distance in the albedo kernel (vs RGB only) (ablation) | single-scan 3DMM · Multi-PIE 45 deg yaw | accuracy: 28.9 -> 54.2 | Improved |
| combine RGB-space and physical-space distance in the albedo kernel (vs XYZ only) (ablation) | single-scan 3DMM · Multi-PIE 45 deg yaw | accuracy: 35.7 -> 54.2 | Improved |
| add bilateral symmetry and color correlation kernels (ablation) | standard-full 3DMM · Multi-PIE 45 deg yaw | accuracy: 54.2 -> 72.3 | Improved |
| add bilateral symmetry and color correlation kernels (ablation) | standard-full 3DMM · Multi-PIE 15 deg yaw | accuracy: 84.7 -> 93.2 | Improved |
| add symmetry on top of color correlation (ablation) | correlated-full 3DMM · Multi-PIE 45 deg yaw | accuracy: 66.7 -> 72.3 | Improved |
| add symmetry on top of color correlation (ablation) | correlated-full 3DMM · Multi-PIE 30 deg yaw | accuracy: 87.1 -> 85.9 | No clear effect |
| use mean of Basel Face Model 2019 as template instead of a single individual scan | symmetric 3DMM · Multi-PIE 45 deg yaw | accuracy: 49.8 -> 72.3 | Improved |
| replace 10-scan PCA with 10-scan KDE mixture of single-scan 3DMMs | 10-scan 3DMM · Multi-PIE 45 deg yaw | accuracy: 42.2 -> 71.5 | Improved |
| replace 10-scan PCA with 10-scan KDE mixture of single-scan 3DMMs | 10-scan 3DMM · Multi-PIE 15 deg yaw | accuracy: 86.0 -> 94.0 | Improved |
| add albedo information to MCMC registration (vs shape only) (ablation) | single-scan 3DMM registration · registration of 10 scans | shape error: slightly increases shape error | Mixed |
| learn 3DMM from 2D images by wake-sleep iterations | standard-full / symmetric-full 3DMM · Multi-PIE face recognition | accuracy | Hurt |
Closest papers by meaning
- Learning Complete 3D Morphable Face Models From Images and Videos (CVPR 2021 · similarity 0.91)
- Generating 3D Faces using Convolutional Mesh Autoencoders (ECCV 2018 · similarity 0.90)
- Unsupervised Training for 3D Morphable Model Regression (CVPR 2018 · similarity 0.89)
- Learning Formation of Physically-Based Face Attributes (CVPR 2020 · similarity 0.89)
- Regressing Robust and Discriminative 3D Morphable Models With a Very Deep Neural Network (CVPR 2017 · similarity 0.89)
- Common3D: Self-Supervised Learning of 3D Morphable Models for Common Objects in Neural Feature Space (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.
