update README

This commit is contained in:
Sanster 2022-03-06 20:43:12 +08:00
parent f9b96cf218
commit 4e5903129c

View File

@ -31,6 +31,11 @@ python3 main.py --device=cuda --port=8080 --model=ldm --ldm-steps=50
Diffusion model is **MUCH MORE** slower than GANs(1080x720 image takes 8s on 3090), but it's possible to get better Diffusion model is **MUCH MORE** slower than GANs(1080x720 image takes 8s on 3090), but it's possible to get better
results than LaMa. results than LaMa.
|Original Image|LaMa|LDM|
|--------------|------|----|
|![photo-1583445095369-9c651e7e5d34](https://user-images.githubusercontent.com/3998421/156923525-d6afdec3-7b98-403f-ad20-88ebc6eb8d6d.jpg)|![photo-1583445095369-9c651e7e5d34_cleanup_lama](https://user-images.githubusercontent.com/3998421/156923620-a40cc066-fd4a-4d85-a29f-6458711d1247.png)|![photo-1583445095369-9c651e7e5d34_cleanup_ldm](https://user-images.githubusercontent.com/3998421/156923652-0d06c8c8-33ad-4a42-a717-9c99f3268933.png)|
Blogs about diffusion models: Blogs about diffusion models:
- https://lilianweng.github.io/posts/2021-07-11-diffusion-models/ - https://lilianweng.github.io/posts/2021-07-11-diffusion-models/