OmniFake is a large-scale, well-categorized synthetic image dataset introduced in Few-Shot Synthetic Image Attribution: Identifying Unseen Generators with Limited Samples. It contains 1.17 million AI-generated images from 45 distinct generators, paired with 1.17 million real images, designed for research on AI-generated image (AIGI) detection and source attribution.
For usage instructions and experimental protocols, please refer to the OmniDFA GitHub repository.… See the full description on the dataset page:
https://huggingface.co/datasets/MoeNew/OmniFake.
