MultiFakeVerse

MultiFakeVerse数据集由莫纳什大学的研究人员创建，包含845,286张通过视觉语言模型（VLM）进行图像感知修改生成的合成图像。该数据集专注于对场景中的人物或上下文元素进行针对性修改，以影响人类对图像重要性的感知。MultiFakeVerse旨在挑战现有的检测方法，加深对感知操纵的理解，并推动对操纵视觉内容的鲁棒检测算法的研究。

The MultiFakeVerse dataset was developed by researchers at Monash University, containing 845,286 synthetic images generated via image perceptual modification using visual-language models (VLMs). This dataset focuses on targeted modifications to human figures or contextual elements within scenes, with the goal of altering human perceptions of the importance of image content. MultiFakeVerse is designed to challenge existing detection methods, deepen the understanding of perceptual manipulation, and advance research on robust detection algorithms for manipulated visual content.