LFW-Beautified

LFW-Beautified数据集是由哈姆斯塔德大学智能系统应用研究中心创建，旨在研究社交媒媒体中常用的美化及增强现实滤镜对人脸检测和识别的影响。该数据集包含34,592张64x64像素的图像，源自流行的Labeled Faces in the Wild (LFW)数据库，通过应用多种图像处理算法模拟社交媒媒体滤镜效果。数据集创建过程中，使用了9种流行的Instagram自拍滤镜和4种增强现实滤镜，以及一个专门训练的U-Net网络来逆向处理太阳镜滤镜。该数据集适用于开发能够抵抗社交媒媒体滤镜影响的人脸检测和识别系统，以及社交媒媒体犯罪调查和图像预处理等应用。

The LFW-Beautified dataset was developed by the Applied Research Center for Intelligent Systems at Halmstad University, aiming to investigate the impact of commonly used beautification and augmented reality (AR) filters on social media on face detection and recognition. This dataset contains 34,592 64x64 pixel images derived from the popular Labeled Faces in the Wild (LFW) database, where various image processing algorithms are applied to simulate the effects of social media filters. During the dataset creation process, 9 popular Instagram selfie filters, 4 AR filters, and a specially trained U-Net network were used to reverse-process the sunglass filter effect. This dataset is suitable for developing face detection and recognition systems robust to the interference of social media filters, as well as applications such as social media crime investigation and image preprocessing.