Research map / Label-efficient and robust learning / Recognition and long-tail
GridFace: Face Rectification via Learning Local Homography Transformations
Erjin Zhou, Zhimin Cao, Jian Sun
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 8x8 local homography rectification (Grid-8) to the recognition baseline | baseline · SNFace test set | verification accuracy at FAR=1e-4: 63.41 -> 72.71 | Improved |
| increase homography cells from global (Grid-1) to 4x4 (Grid-4) to 8x8 (Grid-8) (ablation) | GridFace · SNFace test set | verification accuracy at FAR=1e-4: 66.15 -> 68.70 -> 71.09 -> 72.71 | Improved |
| add face prior regularization to Grid-8 (ablation) | Grid-8 · SNFace test set | verification accuracy at FAR=1e-4: 69.05 -> 72.71 | Improved |
| add face prior regularization to Grid-8 (ablation) | Grid-8 · SNFace test set | verification accuracy at FAR=1e-3: 85.44 -> 87.81 | Improved |
| learn rectification and recognition jointly instead of sequentially (ablation) | GridFace sequential · SNFace test set | verification accuracy at FAR=1e-4: 41.6 -> 72.71 | Improved |
| train recognition on 3D-frontalized faces (baseline-3D) instead of original faces | baseline · SNFace test set | verification accuracy at FAR=1e-4: 63.41 -> 58.20 | Hurt |
| train recognition on 3D-frontalized faces (baseline-3D) instead of original faces | baseline · SNFace test set | verification accuracy at FAR=1e-2: 92.94 -> 94.02 | Improved |
| replace 3D frontalization (baseline-3D) with Grid-8 | baseline-3D · SNFace test set | verification accuracy at FAR=1e-4: 58.20 -> 72.71 | Improved |
| add Grid-8 rectification | Baseline (SNFace training) · LFW | open-set DIR at FAR 1%: 80.3 -> 94.1 | Improved |
| add Grid-8 rectification | Baseline (SNFace training) · LFW | mean accuracy: 99.15 -> 99.70 | Improved |
| add Grid-8 rectification | Baseline (SNFace training) · YTF | mean accuracy: 94.0 -> 95.6 | Improved |
| add Grid-8 rectification | Baseline (SNFace training) · Multi-PIE yaw 90 | identification rate: 65.5 -> 75.4 | Improved |
Closest papers by meaning
- InverseFaceNet: Deep Monocular Inverse Face Rendering (CVPR 2018 · similarity 0.85)
- RegularFace: Deep Face Recognition via Exclusive Regularization (CVPR 2019 · similarity 0.85)
- Self-Supervised Multi-Level Face Model Learning for Monocular Reconstruction at Over 250 Hz (CVPR 2018 · similarity 0.84)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.84)
- Data Synthesis with Diverse Styles for Face Recognition via 3DMM-Guided Diffusion (CVPR 2025 · similarity 0.83)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.83)
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.
