Margin-based deep face recognition methods (e.g. SphereFace, CosFace, and ArcFace) have achieved remarkable success in unconstrained face recognition. However, these methods are susceptible to the massive label noise in the training data and thus require laborious human effort to clean the datasets. In this paper, we relax the intra-class constraint of ArcFace to improve the robustness to label noise. More specifically, we design K sub-centers for each class and the training sample only needs to be close to any of the K positive subcenters instead of the only one positive center. The proposed sub-center ArcFace encourages one dominant sub-class that contains the majority of clean faces and non-dominant sub-classes that include hard or noisy faces. Extensive experiments confirm the robustness of sub-center ArcFace under massive real-world noise. After the model achieves enough discriminative power, we directly drop non-dominant sub-centers and high-confident noisy samples, which helps recapture intra-compactness, decrease the influence from noise, and achieve comparable performance compared to ArcFace trained on the manually cleaned dataset. By taking advantage of the large-scale raw web faces (Celeb500K), sub-center Arcface achieves state-of-the-art performance on IJB-B, IJB-C, MegaFace, and FRVT.

Link to Sub-center arcface: Boosting face recognition by large-scale noisy web faces

Citation:

@inproceedings{deng2020sub,
                    title={Sub-center arcface: Boosting face recognition by large-scale noisy web faces},
                    author={Deng, Jiankang and Guo, Jia and Liu, Tongliang and Gong, Mingming and Zafeiriou, Stefanos},
                    booktitle={European Conference on Computer Vision},
                    pages={741--757},
                    year={2020},
                    organization={Springer}
                  }