ROF

该数据集名为ROF，包含了5559张180位名人的图片，其中每位名人都至少有一张佩戴太阳镜的照片，这些照片被用作语义后门攻击的毒化数据集。此外，为了进行基于纹身和基于面具的攻击，还创建了使用Snapchat滤镜合成的变体。该数据集规模较大，共有5559张图片，其任务旨在人脸识别中的后门攻击防御。

This dataset, named ROF, consists of 5,559 images of 180 celebrities, with each celebrity having at least one photograph where they are wearing sunglasses. These images serve as a poisoning dataset for semantic backdoor attacks. Furthermore, variants synthesized using Snapchat filters were created to support tattoo-based and mask-based attacks. As a large-scale dataset with a total of 5,559 images, it is designed for backdoor attack defense research in facial recognition.