search

xXCoffeeColaXc / DiffusionBasedImageTranlationWithGradientGuidance

This repository hosts an implementation of the Diffusion-based Image Translation with Label Guidance for Domain Adaptive Semantic Segmentation paper. Aimed at translating images from a source domain to a target domain, while keeping the semantic consistency between the domains.
MIT License
1 stars 0 forks source link

How to get same translated image as source without adding any extra or new objects #1

Open IqraNosheen786 opened 2 months ago

IqraNosheen786 commented 2 months ago

Hi, based on the results attached, it is still changing the original architecture of the source image and adding extra objects to the translated image? https://github.com/xXCoffeeColaXc/DiffusionBasedImageTranlationWithGradientGuidance/blob/main/res/to_night.png

xXCoffeeColaXc commented 2 months ago

@IqraNosheen786 Hello, yes it does. I will look into it. By the end of the summer, I think, I will have a solution for translating with almost perfect semantic consistency!

IqraNosheen786 commented 2 months ago

@xXCoffeeColaXc I am also actively exploring the use of diffusion models for domain adaptation tasks. My goal is to make virtual data more realistic while preserving its original structure. Looking forward to your solution.