Community Forensics: Using Thousands of Generators to Train Fake Image Detectors

2024-11-06

arXiv.org

AI Summary

Existing AI-generated image detectors exhibit insufficient generalization to unseen text-to-image diffusion models. Method: We introduce the largest forgery image benchmark to date—comprising 4,803 distinct diffusion models and 2.7 million highly diverse images—uniquely integrating both open-source and commercial models across architectures and training configurations. We propose an end-to-end framework encompassing automated web crawling, heterogeneous model integration, and unified detector training. Contribution/Results: Empirical analysis reveals that both model count and architectural diversity confer dual positive gains in detector generalization. Our approach achieves significantly higher detection accuracy on previously unseen generators than current state-of-the-art methods, empirically validating data diversity as an effective pathway to enhanced generalization in AI-generated image detection.

Technology Category

Computer Vision: Diffusion Models for VisionMachine Learning: Large Multimodal Models (LMMs)Natural Language Processing: Generation

Application Category

Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphsWeb Mining and Content Analysis: Large pretrained models with web dataSearch and Retrieval-Augmented AI: Web search models and ranking

Abstract

One of the key challenges of detecting AI-generated images is spotting images that have been created by previously unseen generative models. We argue that the limited diversity of the training data is a major obstacle to addressing this problem, and we propose a new dataset that is significantly larger and more diverse than prior work. As part of creating this dataset, we systematically download thousands of text-to-image latent diffusion models and sample images from them. We also collect images from dozens of popular open source and commercial models. The resulting dataset contains 2.7M images that have been sampled from 4803 different models. These images collectively capture a wide range of scene content, generator architectures, and image processing settings. Using this dataset, we study the generalization abilities of fake image detectors. Our experiments suggest that detection performance improves as the number of models in the training set increases, even when these models have similar architectures. We also find that detection performance improves as the diversity of the models increases, and that our trained detectors generalize better than those trained on other datasets.

Problem

• Detecting AI-generated images from unseen models
• Overcoming limited training data diversity
• Improving fake image detector generalization

Innovation

• Larger and more diverse training dataset
• Thousands of text-to-image diffusion models
• Improved fake image detection generalization