Replies: 2 comments 9 replies
-
@huchenlei大佬,求助,我已经下载了1.5G的深度模型改名后放入了F:\sd-webui-aki-v4.5\extensions\sd-webui-controlnet\models 文件夹里面 |
Beta Was this translation helpful? Give feedback.
9 replies
-
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Depth Anything project
https://github.com/LiheYoung/Depth-Anything
How to use
Download depth_anything ControlNet model here. You need to rename the file for ControlNet extension to correctly recognize it. We recommend user to rename it as
control_sd15_depth_anything
.Model comparison
Depth anything comes with a preprocessor and a new SD1.5 ControlNet model trained with images annotated by this preprocessor. They do not need to be used together. Here is a grid of result comparisons against existing preprocessors/models.
input image

midas

leres

leres++

depth_anything

xyz grid

Conslusion
I would say depth_anything might not outperforms depth_leres++ for amount details captured. We still adds the preprocessor anyway as it is fairly lightweight code wise.
Update 2024-01-23
According to thread https://www.reddit.com/r/StableDiffusion/comments/19d7fx1/controlnet_has_been_updated_to_1435_includes/, depth_anything is better at handling motion blur and has better temporal consistency/stability. Maybe it is a superior choice when doing video generation, i.e. AnimateDiff.
Beta Was this translation helpful? Give feedback.
All reactions