Research map / 3D vision / Face and shape
A Hierarchical Representation Network for Accurate and Detailed Face Reconstruction From In-the-Wild 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 hierarchical representation modeling (ablation) | HRN · FaceScape-wild | CD (mm): 3.8 -> 3.34 | Helped |
| use hierarchical representation modeling (ablation) | HRN · FaceScape-wild | MNE (rad): 0.092 -> 0.081 | Helped |
| add contour-aware loss (ablation) | HRN with hierarchical modeling · FaceScape-wild | CD (mm): 3.34 -> 3.18 | Helped |
| add contour-aware loss (ablation) | HRN with hierarchical modeling · FaceScape-wild | MNE (rad): 0.081 -> 0.079 | Helped |
| add 3D detail priors (ablation) | HRN with contour loss · FaceScape-wild | CD (mm): 3.18 -> 2.9 | Helped |
| add 3D detail priors (ablation) | HRN with contour loss · FaceScape-wild | MNE (rad): 0.079 -> 0.067 | Helped |
| add de-retouching module (ablation) | HRN with detail priors · FaceScape-wild | CD (mm): 2.9 -> 2.91 | Hurt |
| add de-retouching module (ablation) | HRN with detail priors · FaceScape-wild | MNE (rad): 0.067 -> 0.065 | Helped |
| use hierarchical modeling and detail priors (ablation) | HRN · FaceScape-wild | CD (mm): The hierarchical modeling strategy brings a huge improvement ($\sim$0.5mm); 3D priors achieve $\sim$0.3mm improvement. | Helped |
| increase input views from 2 to 5 (ablation) | MV-HRN · FaceScape | Median distance (mm): 1.17 -> 1.10 | Helped |
| replace the multi-view design with naive averaging (ablation) | MV-HRN · FaceScape (5 views) | Median distance (mm): 1.10 -> 1.23 | Hurt |
| use HRN instead of the strongest listed baseline, 3DDFA-V2 | single-view face reconstruction · FaceScape-wild | CD (mm): 3.00 -> 2.91 | Helped |
Closest papers by meaning
- Towards High-Fidelity 3D Face Reconstruction From In-the-Wild Images Using Graph Convolutional Networks (CVPR 2020 · similarity 0.90)
- Photorealistic Facial Texture Inference Using Deep Neural Networks (CVPR 2017 · similarity 0.89)
- Detailed Facial Geometry Recovery from Multi-View Images by Learning an Implicit Function (AAAI 2022 · similarity 0.88)
- Extreme 3D Face Reconstruction: Seeing Through Occlusions (CVPR 2018 · similarity 0.88)
- MMFace: A Multi-Metric Regression Network for Unconstrained Face Reconstruction (CVPR 2019 · similarity 0.88)
- Self-Supervised Multi-Level Face Model Learning for Monocular Reconstruction at Over 250 Hz (CVPR 2018 · similarity 0.88)
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.