Using stable diffusion and segmentation anything models for image editing.
Generally, given a textual prompt or cliked region, SAM generated the masked region for source image. Then, we use CLIP model to select the region, which can be finally used to generate the target edited image with stable diffusion.
Use python service.py to initialize the service.
Moreover, we introduce a more advanced inpainting model from diffusers to suport better image editing. Use python service_img2img.py to launch the service.
[1] https://github.com/huggingface/diffusers
[2] https://github.com/facebookresearch/segment-anything
[3] https://github.com/maxi-w/CLIP-SAM
[4] https://github.com/IDEA-Research/Grounded-Segment-Anything/