wyhuai / DDNM

[ICLR 2023 Oral] Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model
MIT License
1.16k stars 83 forks source link

Zero-Shot Image Restoration model for final denoising steps in Image Generation task? #49

Open 5thGenDev opened 1 year ago

5thGenDev commented 1 year ago

Let's say an Image Generation model is meant to generate image from random noise in 50 steps:

  1. What if I use Image Generation model for the initial 20 steps to generate a half-way noisy image.
  2. Input that noisy image into a Zero-Shot Image Restoration model to generate a clean image.

Has anyone tried this approach? Does it have any advantage in classical image generation quality, few-shot image generation?

wyhuai commented 1 year ago

I haven't tried it, but I think this probably won't produce better quality than the default method.

5thGenDev commented 1 year ago

fair enough, this will be the topic for my thesis then :P

wyhuai commented 1 year ago

Good luck and hope you can find a way to make it work!