Diffusion with Image Forensics.

[CHELSEA234]

Hi @heejkoo , i think we can include diffusion models from the image forensic perspective, which states how to distinguish image generated by diffusion model from real images. This is a meaningful research direction and has many practical needs in the security system, as well as helps people use diffusion model to generate more "real" images. This image forensic has many topics, such as localization, detection, and attribution. Please consider taking a look on these following works:

Towards the Detection of Diffusion Model Deepfakes (https://arxiv.org/pdf/2210.14571.pdf) Hierarchical Fine-Grained Image Forgery Detection and Localization (CVPR2023) (https://arxiv.org/pdf/2303.17111.pdf) AutoSplice: A Text-prompt Manipulated Image Dataset for Media Forensics (https://arxiv.org/pdf/2304.06870.pdf) DE-FAKE: Detection and Attribution of Fake Images Generated by Text-to-Image Generation Models (https://arxiv.org/pdf/2210.06998.pdf)

[diff-usion]

Hello, I will search for related works and add the "image forensic" category sooner or later. Thank you for your valuable opinion.

[CHELSEA234]

Hello, I will search for related works and add the "image forensic" category sooner or later. Thank you for your valuable opinion.

@heejkoo May I ask can we add this new direction in the github now?

[diff-usion]

Sorry for being late. I will add the new subsection sooner or later. Thank you for your patience and understanding!