Large-scale Labeled Face (LSLF) 和 noisy Large-scale Labeled Non-face (LSLNF)

Large-scale Labeled Face (LSLF) 数据集由韦恩州立大学计算机科学系创建，包含1,195,976张多视角和部分遮挡的面部图像，涵盖多种颜色、灰度、图像质量、分辨率、光照、背景等变化。数据集通过自动化的YouTube视频链接爬虫、下载器和Viola-Jones面部检测等五阶段流程创建，旨在为大规模面部学习和识别任务提供丰富的训练样本。LSLF是目前文献中最大的标记面部图像数据集，适用于解决面部识别中的复杂场景问题。

Large-scale Labeled Face (LSLF) dataset was created by the Department of Computer Science at Wayne State University. It contains 1,195,976 facial images with multiple viewpoints and partial occlusions, encompassing variations in color, grayscale, image quality, resolution, illumination, background and other aspects. The dataset is constructed through a five-stage pipeline including automated YouTube video link crawling, downloading, Viola-Jones face detection and other procedures, aiming to provide abundant training samples for large-scale facial learning and recognition tasks. LSLF is currently the largest labeled facial image dataset in academic literature, which is suitable for solving complex scenario problems in face recognition.