Fair Forgery Detection (FairFD) dataset
收藏资源简介:
Fair Forgery Detection (FairFD)数据集由西安电子科技大学和重庆邮电大学创建,旨在评估面部伪造检测中的种族偏见。该数据集包含11430个主题,种族比例均衡,涵盖多种伪造生成图像,包括FaceSwap、SimSwap等多种方法。数据集的创建过程注重种族平衡和伪造方法的多样性,适用于评估面部伪造检测系统的公平性,旨在解决现有数据集在种族公平评估方面的不足。
Fair Forgery Detection (FairFD) dataset was developed by Xidian University and Chongqing University of Posts and Telecommunications, targeting the evaluation of racial bias in facial forgery detection. This dataset consists of 11,430 subjects with a well-balanced racial distribution, covering a wide range of forged images generated by diverse methods including FaceSwap, SimSwap and other similar techniques. The construction of FairFD prioritizes both racial balance and the diversity of forgery generation approaches, making it suitable for assessing the fairness of facial forgery detection systems, and it is intended to address the shortcomings of existing datasets in the context of racial fairness evaluation.




