We clean, merge, and release the largest and cleanest face recognition dataset Glint360K, which contains 17091657 images of 360232 individuals. By employing the Patial FC training strategy, baseline models trained on Glint360K can easily achieve state-of-the-art performance. Detailed evaluation results on the large-scale test set (e.g. IFRT, IJB-C and Megaface) are as follows:

Dataset Details… See the full description on the dataset page: https://huggingface.co/datasets/yayoimizuha/Glint360k.