DeepFaceGen
收藏资源简介:
DeepFaceGen是由浙江大学开发的一个大规模人脸伪造检测评估基准。该数据集包含463,583张真实人脸图像和313,407个真实视频,以及350,264张伪造图像和423,548个伪造视频,这些伪造样本使用了34种主流的人脸生成技术。在构建过程中,DeepFaceGen考虑了内容多样性、种族公平性和全面的标签可用性,确保了其多功能性和便利性。该数据集主要用于评估和分析现有面部伪造检测技术,旨在推动面部伪造检测技术的发展,解决由AI生成内容技术引发的真实性验证难题。
DeepFaceGen is a large-scale face forgery detection and evaluation benchmark developed by Zhejiang University. This dataset contains 463,583 real face images, 313,407 real videos, 350,264 forged images and 423,548 forged videos, with the forged samples generated using 34 mainstream face generation technologies. During its construction, DeepFaceGen took into account content diversity, racial fairness and comprehensive label availability, ensuring its versatility and convenience. This dataset is mainly used to evaluate and analyze existing face forgery detection technologies, aiming to promote the development of face forgery detection technologies and address the authenticity verification challenges caused by AI-generated content technologies.




