Can I read Fine-Tuning Impact on Fake Image Detection on EtoBox?

Fine-Tuning Impact on Fake Image Detection by Mathan is a document available to read on EtoBox.

What is Fine-Tuning Impact on Fake Image Detection about?

This thesis investigates the impact of fine-tuning methods, DreamBooth and LoRA, on the performance of fake image detectors trained on Stable Diffusion models. The study finds that detectors trained solely on base model images perform worse when tested on images generated by fine-tuned models, with DreamBooth having a more significant negative effect on accuracy than LoRA. The results highlight the necessity to treat fine-tuned models as distinct entities for effective fake image detection training.

Author

Mathan

Language

EN