ARIA

ARIA数据集是由堪萨斯大学电气工程与计算机科学系/智能系统研究所创建，包含超过14万张图像，分为五个类别：艺术品、社交媒体图像、新闻照片、灾难场景和动漫图片。该数据集通过四个流行的AI图像生成器生成，旨在支持对抗性AI艺术的研究。数据集创建过程涉及使用文本描述作为提示，生成与真实图像对应的AI图像。ARIA数据集的应用领域包括检测和理解AI生成的图像，以及评估人类和AI系统识别这些图像的能力，旨在解决AI生成图像在现实世界中可能引发的安全和伦理问题。

The ARIA Dataset was developed by the Department of Electrical Engineering and Computer Science / Institute for Intelligent Systems, University of Kansas. It contains over 140,000 images categorized into five groups: artwork, social media images, news photos, disaster scenes, and anime images. This dataset was generated using four mainstream AI image generators, and is specifically designed to support research on adversarial AI art. The dataset creation workflow involved using text descriptions as prompts to produce AI-generated images that match real-world images. The ARIA Dataset can be applied to detect and interpret AI-generated images, as well as evaluate the performance of both humans and AI systems in identifying such images. Its core goal is to address the security and ethical concerns posed by AI-generated images in real-world scenarios.