OwensLab/CommunityForensics-Small

Community Forensics (small) 数据集旨在开发和应用图像取证方法，用于检测或分析 AI 生成的图像。该数据集包含从 4803 个生成器模型中收集的约 278,000 张生成图像，以及来自 FFHQ、VISION、COCO 和 Landscapes HQ 数据集的 278,000 张真实图像。数据集以 Parquet 格式组织，并提供图像名称、格式、分辨率、模式、图像数据、模型名称、NSFW 标志、提示、真实来源、子集、分割、标签和架构等字段信息。数据集可用于图像分类任务，识别图像是否为 AI 生成。数据集的默认分割包含配对的训练数据集，包含 278,000 张真实图像和 278,000 张生成图像。

The Community Forensics (small) dataset is designed for developing and benchmarking forensic methods to detect or analyze AI-generated images. It contains approximately 278,000 generated images collected from 4803 generator models, paired with 278,000 real images sourced from FFHQ, VISION, COCO, and Landscapes HQ datasets. The dataset is structured in a Parquet data frame and includes fields such as image name, format, resolution, mode, image data, model name, NSFW flag, prompt, real source, subset, split, label, and architecture. It is suitable for image classification tasks to determine whether an image is AI-generated. The default split provides a paired training dataset with 278,000 real and 278,000 generated images.