Research map / 3D vision / Face and shape
REALY: Rethinking the Evaluation of 3D Face Reconstruction
Zenghao Chai, Haoxian Zhang, Jing Ren, Di Kang, Zhengzhuo Xu, Xuefei Zhe, Chun Yuan, Linchao Bao
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 |
|---|---|---|---|
| use region-wise bICP instead of global gICP for correspondence (ablation) | evaluation pipeline · nn-map error, nose | error (mm): 2.882 -> 0.670 | Improved |
| use region-wise bICP instead of global gICP for correspondence (ablation) | evaluation pipeline · nn-map error, mouth | error (mm): 1.699 -> 0.407 | Improved |
| use region-wise bICP instead of global gICP for correspondence (ablation) | evaluation pipeline · nn-map error, forehead | error (mm): 0.707 -> 0.520 | Improved |
| use region-wise bICP instead of global gICP for correspondence (ablation) | evaluation pipeline · nn-map error, cheek | error (mm): 1.219 -> 0.105 | Improved |
| add bidirectional step to region-wise rICP (ablation) | evaluation pipeline · nn-map error, nose | error (mm): 0.706 -> 0.670 | Improved |
| add bidirectional step to region-wise rICP (ablation) | evaluation pipeline · nn-map error, cheek | error (mm): 0.107 -> 0.105 | No clear effect |
| align with region-wise rICP instead of gICP (ablation) | evaluation pipeline · alignment distance to S_L, nose-modified shape, nose region | e (mm/10): 40.490 -> 8.376 | Improved |
| evaluate with bICP instead of gICP | face reconstruction methods · REALY | best-method ranking | Mixed |
| use HIFI3D^A instead of HIFI3D basis in RGB-D fitting | 3DMM fitting · REALY RGB-D fitting | bICP error all (mm): 1.093 -> 0.972 | Improved |
| use the new 3DMM basis instead of HIFI3D^A | 3DMM fitting · REALY RGB-D fitting | bICP error all (mm): 0.972 -> 0.878 | Improved |
| use RGB-D instead of RGB fitting | HIFI3D^A · REALY | bICP error all (mm): 2.167 -> 0.972 | Improved |
| use RGB-D instead of RGB fitting | FS · REALY | bICP error all (mm): 2.386 -> 1.927 | Improved |
Closest papers by meaning
- Discovering Geometric Biases in 3D Face Reconstruction: A Curvature-Aware Spectral Framework for Fairness Evaluation (ECCV 2026 · similarity 0.89)
- How Far Are We From Solving the 2D & 3D Face Alignment Problem? (And a Dataset of 230,000 3D Facial Landmarks) (ICCV 2017 · similarity 0.88)
- Faster Than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses (ICCV 2017 · similarity 0.88)
- Topologically Consistent Multi-View Face Inference Using Volumetric Sampling (ICCV 2021 · similarity 0.88)
- FaceScape: 3D Facial Dataset and Benchmark for Single-View 3D Face Reconstruction (TPAMI 2023 · similarity 0.88)
- Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network (ECCV 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.
