DeepFaceGen
收藏资源简介:
DeepFaceGen是由浙江大学构建的大型人脸伪造检测评估基准,旨在量化评估人脸伪造检测技术的有效性,并促进伪造检测技术的迭代发展。该数据集包含776,990个真实人脸图像/视频样本和773,812个人脸伪造图像/视频样本,这些伪造样本使用了34种主流人脸生成技术生成。在构建过程中,DeepFaceGen考虑了内容多样性、种族公平性和全面标签的可获取性,以确保其多功能性和便利性。该数据集主要应用于人脸伪造检测领域,旨在解决由AI生成内容技术快速发展带来的真实性验证难题,增强多媒体信息的信任度,并降低社会安全风险。
DeepFaceGen is a large-scale benchmark for face forgery detection and evaluation developed by Zhejiang University, which is designed to quantitatively evaluate the effectiveness of face forgery detection technologies and facilitate the iterative advancement of such techniques. This dataset contains 776,990 real face image/video samples and 773,812 forged face image/video samples, which were generated via 34 mainstream face generation technologies. During its construction, DeepFaceGen took into account content diversity, racial fairness, and the accessibility of comprehensive annotations to ensure its versatility and convenience. This dataset is primarily applied in the field of face forgery detection, aiming to address the authenticity verification challenges brought about by the rapid development of AI-generated content technologies, enhance the credibility of multimedia information, and mitigate social security risks.




