Research map / 3D vision / Face and shape
Graphics Capsule: Learning Hierarchical 3D Face Representations From 2D Images
Read the paperCheck 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 graphics capsules with hierarchical 3D representations instead of a 2D autoencoder representation | Face recognition · Multi-PIE cross-view face recognition | recognition accuracy: 91.2\% -> 95.1\% | Helped |
| Use IGC-Net instead of SCOPS | Unsupervised face segmentation · CelebA | NME_L (%): 18.72 -> 11.84 | Helped |
| Use IGC-Net instead of SCOPS | Unsupervised face segmentation · CelebA | NME_DL (%): 23.69 -> 18.88 | Helped |
| Use IGC-Net instead of HP-Capsule | Unsupervised face segmentation · BP4D | NME_L (%): 8.81 -> 6.35 | Helped |
| Use IGC-Net instead of HP-Capsule | Unsupervised face segmentation · BP4D | NME_DL (%): 6.10 -> 4.32 | Helped |
| Add the one-hot operation to GDM (ablation) | IGC-Net · CelebA | NME_L: 19.10 -> 11.84 | Helped |
| Add the contrastive semantic constraint (ablation) | IGC-Net · CelebA | NME_L: 13.46 -> 11.84 | Helped |
| Remove the one-hot operation from GDM (ablation) | IGC-Net · CelebA | semantic consistency of discovered parts | Hurt |
| Remove the contrastive semantic constraint (ablation) | IGC-Net · CelebA | salient part discovery | Hurt |
Closest papers by meaning
- HP-Capsule: Unsupervised Face Part Discovery by Hierarchical Parsing Capsule Network (CVPR 2022 · similarity 0.90)
- InverseFaceNet: Deep Monocular Inverse Face Rendering (CVPR 2018 · similarity 0.86)
- 3D-RCNN: Instance-Level 3D Object Reconstruction via Render-and-Compare (CVPR 2018 · similarity 0.86)
- Learning Unsupervised Hierarchical Part Decomposition of 3D Objects From a Single RGB Image (CVPR 2020 · similarity 0.85)
- 3DP3: 3D Scene Perception via Probabilistic Programming (NeurIPS 2021 · similarity 0.85)
- PolyhedronNet: Representation Learning for Polyhedra with Surface-attributed Graph (ICLR 2025 · similarity 0.84)
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.