Research map / Label-efficient and robust learning / Noisy labels
Global-Local GCN: Large-Scale Label Noise Cleansing for Face Recognition
Yaobin Zhang, Weihong Deng, Mei Wang, Jiani Hu, Xian Li, Dongyue Zhao, Dongchao Wen
Check 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 a local GCN stage to refine global predictions (ablation) | FaceGraph · IMDb-Face simulation, garbage 10%, outliers 30%, label flips 30% | F (%): 89.09 -> 90.03 | Improved |
| Train global and local networks separately rather than cooperatively (ablation) | FaceGraph · IMDb-Face simulation, garbage 10%, outliers 30%, label flips 30% | F (%): 90.03 -> 89.86 | Hurt |
| Remove multi-task learning from cooperative training (ablation) | FaceGraph · IMDb-Face simulation, garbage 10%, outliers 30%, label flips 30% | F (%): 89.47 -> 90.03 | Hurt |
| Cleanse MS1M with FaceGraph instead of DeepGlint | Arcface · MS1M, average of seven verification sets | Average accuracy (%): 95.32 -> 96.78 | Improved |
| Train on MillionCelebs cleansed with FaceGraph instead of MegaFace2 | ResNet100 · MegaFace Challenge 1 | Identification Rank-1 (%): 98.96 -> 99.02 | Improved |
| Train on MillionCelebs cleansed with FaceGraph instead of MegaFace2 | ResNet100 · IJB-C | Verification TPR (@FPR=1e-5) (%): 76.29 -> 95.62 | Improved |
Closest papers by meaning
- The Devil of Face Recognition is in the Noise (ECCV 2018 · similarity 0.90)
- Sub-center ArcFace: Boosting Face Recognition by Large-scale Noisy Web Faces (ECCV 2020 · similarity 0.87)
- Adaptive Label Noise Cleaning With Meta-Supervision for Deep Face Recognition (ICCV 2021 · similarity 0.87)
- RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition (AAAI 2025 · similarity 0.87)
- Noise-Robust Face Recognition via Non-target Similarity Distribution Guided Sample Selection (ECCV 2026 · similarity 0.86)
- NoisyGL: A Comprehensive Benchmark for Graph Neural Networks under Label Noise (NeurIPS 2024 · 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.
