search

Shilin-LU / TF-ICON

[ICCV 2023] "TF-ICON: Diffusion-Based Training-Free Cross-Domain Image Composition" (Official Implementation)
https://shilin-lu.github.io/tf-icon.github.io/
MIT License
787 stars 102 forks source link

About composition on x-ray dataset #26

Closed pILLOW-1 closed 4 days ago

pILLOW-1 commented 5 days ago

Hi, great work! It is known that there is a gap between x-ray and natural image domains. I wonder if TF-ICON can do image composition on x-ray image datasets. Thanks a lot!

Shilin-LU commented 4 days ago

Thank you for your interest!

TF-ICON is a training-free method, and its capabilities are limited by the pretrained model it relies on. We developed it using Stable Diffusion, which is designed to handle only natural or real images. Therefore, TF-ICON is not suitable for processing X-ray images.

pILLOW-1 commented 4 days ago

Got it. Thank you!