Segment Anything for Stable Diffusion WebUI. Automatically generate high-quality segmentations/masks for images by clicking or text prompting. Aim for connecting WebUI and ControlNet with Segment Anything and GroundingDINO to enhance Stable Diffusion/ControlNet inpainting (both single image and batch process), enhance ControlNet semantic segmentation, automate image matting and create LoRA/LyCORIS training set.

segment-anything stable-diffusion stable-diffusion-webui stable-diffusion-webui-plugin

Go to file

Chengsong Zhang 63b69324bc runnable version		2023-04-13 05:36:01 +08:00
.github/ISSUE_TEMPLATE	add issue template	2023-04-10 22:20:15 +08:00
javascript	usable groundingdino	2023-04-12 11:13:28 +08:00
models	runnable version	2023-04-13 05:36:01 +08:00
scripts	runnable version	2023-04-13 05:36:01 +08:00
.gitignore	update installment	2023-04-12 12:17:54 +08:00
README.md	runnable version	2023-04-13 05:36:01 +08:00
install.py	runnable version	2023-04-13 05:36:01 +08:00
requirements.txt	runnable version	2023-04-13 05:36:01 +08:00
style.css	fix firefox problem	2023-04-11 15:50:58 +08:00

README.md

Segment Anything for Stable Diffusion WebUI

This developing branch attempts to enable GroundingDINO, achieving the goal of text prompt->object detection->segment anything->stable diffusion inpainting.

Warning: You may need a long time to get GroundingDINO and its dependency build, because they have C++.

To use this extension, simply run git checkout GroundingDINO on your terminal after cd to this extension directory. This branch will be merged to master branch someday, when I think it is stable enough.

Next Step:

point + detection, find a best way to satisfy need:
- 0 point + N detection->box: BHW, mask: BCHW
- ALL point + 1 detection->3 images
- Preview detection + select the best

How to Use

TODO

Demo

TODO