-
Notifications
You must be signed in to change notification settings - Fork 4
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
The parameter settings for SD reconstruction #1
Comments
Thank you for your interest in our paper.“SDv1-DR”, “SDv2-DR”, and “SDXL-DR” are utilized to generate reconstruction images of the real images using the SD inpainting models (such as " runwayml/ stable-diffusion-inpainting"), with input conditions being the image, text prompt, and binary mask, where the input image comes from the real image in MSCOCO, the prompt is an empty string “”, and the mask is a zero matrix of the same size as the real image.
Moreover, the values of guidance scale and other parameters are using the default values in diffusers.StableDiffusionInpaintPipeline or diffusers. AutoPipelineForInpainting.
…---Original---
From: ***@***.***>
Date: Thu, Jul 4, 2024 10:46 AM
To: ***@***.***>;
Cc: ***@***.***>;
Subject: [beibuwandeluori/DRCT] The parameter settings for SD reconstruction(Issue #1)
Great work! Could you please share the details of your parameter settings when using the SD model for reconstruction? For example, the prompt, denoising strength and guidance scale?
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>
|
Sorry if my previous question caused any confusion. What I meant was whether the paper used the above "runwayml/stable-diffusion-inpainting" and its corresponding default parameters for reconstruction during training on gen-image(sd1.4)? Thank you very much for your answer! |
Because in my understanding, if the mask a zero matrix, it actually only goes through the VAE process, and the noising/denoising process does not participate at all; while if the mask is all ones and uses the default parameters, the denoising result will be completely independent of the input image. I feel that both of these settings are quite confusing. |
Great work! Could you please share the details of your parameter settings when using the SD model for reconstruction? For example, the prompt, denoising strength and guidance scale?
The text was updated successfully, but these errors were encountered: