Generalizable Origin Identification for Text-Guided Image-to-Image Diffusion Models
Wenhao Wang, Yifan Sun, Zongxin Yang, Zhentao Tan, Zhengdong Hu, Yi Yang
2025-01-08
Summary
This paper talks about a new method to identify the original image used in AI-generated or modified images, especially when the modifications are guided by text instructions.
What's the problem?
Text-guided image-to-image diffusion models are powerful AI tools that can change images based on text descriptions. While useful, these tools can be misused to spread fake information, violate copyrights, or hide the source of images. It's hard to trace back to the original image, especially when different AI models are used.
What's the solution?
The researchers created a new task called ID^2 (Origin IDentification for text-guided Image-to-image Diffusion models) to find the original image from a modified one. They made a special dataset called OriPID with lots of original images and text prompts used to change them. They also developed a method that uses math to find a connection between the modified image and the original, which works across different AI models.
Why it matters?
This matters because as AI-generated images become more common, we need ways to trace their origins. This could help fight fake news, protect copyrights, and make sure people are using AI image tools responsibly. The fact that this method works across different AI models makes it especially useful in real-world situations where we don't always know which AI tool was used to create an image.
Abstract
Text-guided image-to-image diffusion models excel in translating images based on textual prompts, allowing for precise and creative visual modifications. However, such a powerful technique can be misused for spreading misinformation, infringing on copyrights, and evading content tracing. This motivates us to introduce the task of origin IDentification for text-guided Image-to-image Diffusion models (ID^2), aiming to retrieve the original image of a given translated query. A straightforward solution to ID^2 involves training a specialized deep embedding model to extract and compare features from both query and reference images. However, due to visual discrepancy across generations produced by different diffusion models, this similarity-based approach fails when training on images from one model and testing on those from another, limiting its effectiveness in real-world applications. To solve this challenge of the proposed ID^2 task, we contribute the first dataset and a theoretically guaranteed method, both emphasizing generalizability. The curated dataset, OriPID, contains abundant Origins and guided Prompts, which can be used to train and test potential IDentification models across various diffusion models. In the method section, we first prove the existence of a linear transformation that minimizes the distance between the pre-trained Variational Autoencoder (VAE) embeddings of generated samples and their origins. Subsequently, it is demonstrated that such a simple linear transformation can be generalized across different diffusion models. Experimental results show that the proposed method achieves satisfying generalization performance, significantly surpassing similarity-based methods (+31.6% mAP), even those with generalization designs.