search

AUTOMATIC1111 / stable-diffusion-webui

Stable Diffusion web UI
GNU Affero General Public License v3.0
136.27k stars 25.96k forks source link

[Feature Request]: It would be cool to have DIS (Dichotomous Image Segmentation) working inside of your WebUI #7530

Closed starush closed 11 months ago

starush commented 1 year ago

Is there an existing issue for this?

What would your feature do ?

This is Highly Accurate Dichotomous Image Segmentation https://github.com/xuebinqin/DIS#readme It looks like real winner when you want to remove background automatically and with good quality (existing ABG plugin is very basic and is not accurate even with simple shadow). I tested the DIS here (https://huggingface.co/spaces/ECCV2022/dis-background-removal) and found it very useful. It capable separate the main object from really complex images. Impressive!

Proposed workflow

1. I would be happy to see it working in extra tab with batch option, next to upscale. Thank you in advance. 👍

Additional information

No response

Dekker3D commented 1 year ago

Agreed. I'd love to be able to run this on my own PC easily. While it isn't exactly directly related to using Stable Diffusion, this web-UI has a lot of good general purpose stuff in it already. That said, I can imagine that might actually be an argument for putting it in an extension rather than the main repo.

SethRobinson commented 1 year ago

Can agree, it's very useful. I hacked it into my a1111 front-end a few months ago. It does require a 175 MB model download, but hey, that's tiny in the scheme of things. There is supposedly a new improved model for it "coming soon" but it's been delayed for a while now.

catboxanon commented 11 months ago

Supported in https://github.com/AUTOMATIC1111/stable-diffusion-webui-rembg. Closing.