WebFace260M
收藏资源简介:
WebFace260M是一个包含2.6亿张人脸图像的大规模数据集,旨在推动深度人脸识别技术的发展。该数据集由清华大学等机构的研究人员创建,通过自动化的清理流程(CAST)净化数据,确保高质量的训练数据。数据集涵盖了400万身份,适用于多种人脸识别任务,包括标准、带口罩和无偏见的人脸识别。此外,数据集还支持时间约束评估协议(FRUITS),以模拟实际应用场景中的识别挑战。
WebFace260M is a large-scale dataset consisting of 260 million face images, designed to advance the development of deep face recognition technologies. Created by researchers from Tsinghua University and other institutions, this dataset uses an automated cleaning pipeline named CAST to purify the data and ensure high-quality training samples. It covers 4 million distinct identities and supports multiple face recognition tasks, including standard, mask-wearing, and bias-free face recognition. In addition, the dataset provides a time-constrained evaluation protocol (FRUITS) to simulate recognition challenges in real-world application scenarios.




