XavierXiao / Dreambooth-Stable-Diffusion

Implementation of Dreambooth (https://arxiv.org/abs/2208.12242) with Stable Diffusion
MIT License
7.61k stars 795 forks source link

Please post a sample of training data (imgs+regularized) #139

Open SuroshAhmadZobair opened 1 year ago

SuroshAhmadZobair commented 1 year ago

Hi

Thanks for your great repo. I have been struggling to understand what is a regularized image, how to generate it, and what should be the filename of those files.

In the README file, it mentions that we can generate regularized images using txt2img.py but the result does not look good. I wan to generate 2D Floorplan images.

Can you please refer me to a dataset where both the images and the regularized images exist?

Thanks in advance

SuroshAhmadZobair commented 1 year ago

correct me if I am wrong, please!

txt2img.py will generate random images from the pre-trained model based on the prompt I pass. Each image will also have a text alongside the prompt as context. The regularized images will not look 100% like my input images.

Once this is done, I can train the model.