DeepfakeArt Challenge
收藏arXiv2024-05-23 更新2024-06-21 收录
下载链接：
http://anon_for_review.com
下载链接
链接失效反馈官方服务：
资源简介：
DeepfakeArt Challenge是一个专为生成AI艺术伪造和数据中毒检测设计的大型基准数据集，由滑铁卢大学视觉与图像处理研究组创建。该数据集包含超过32,000条记录，涵盖多种生成伪造和数据中毒技术。每个记录由一对图像组成，这些图像可能是伪造的或受到对抗性污染。数据集的创建旨在帮助开发机器学习算法，以识别和防止生成AI在视觉内容合成中的恶意使用，特别是在版权侵犯和数据中毒方面。
The DeepfakeArt Challenge is a large-scale benchmark dataset dedicated to the detection of generative AI art forgeries and data poisoning, developed by the Vision and Image Processing Research Group at the University of Waterloo. This dataset comprises over 32,000 records covering a diverse range of generative forgery and data poisoning techniques. Each record consists of a pair of images that may either be forged or subjected to adversarial contamination. The dataset was constructed to facilitate the development of machine learning algorithms for identifying and mitigating the malicious utilization of generative AI in visual content synthesis, particularly in the contexts of copyright infringement and data poisoning.
提供机构：
滑铁卢大学视觉与图像处理研究组
创建时间：
2023-06-02
