AUTOMATIC1111 / stable-diffusion-webui-rembg

Removes backgrounds from pictures. Extension for webui.
MIT License
1.16k stars 173 forks source link

What is the API for sd webui? #20

Open zhuchenye opened 1 year ago

zhuchenye commented 1 year ago

I want to use the API on webui server with the payload as : payload = { "image": encode_pil_to_base64(image), "model": "u2net", "return_mask": False, "alpha_matting": False, "alpha_matting_foreground_threshold": 240, "alpha_matting_background_threshold": 10, "alpha_matting_erode_size": 10, } but it dos not work.

Is there any docs for remby API?

Theroniya commented 1 year ago

Me too. As a workaround I installed rembg standalone. But why load and wait every time when it is already loaded on the sd server anyway?

zhuchenye commented 1 year ago

I found there is no API route for rembg. So if I used extra_single_image interface, the rembg params will be none.

hubcart commented 1 year ago

It used to work up until a couple weeks ago. I think an update on the webui side broke it. I still have it working on our older setups but any new setups are broke. when working, you have to put postprocessingrembg before each parameter (see pic)

Screenshot 2023-05-25 135050

That setup 100% was working up until about a week or two ago, so if anyone wants to play with that to see what you can find, please post back here if you solve it.

ccfco commented 1 year ago

One alternative approach is to create a new file called api.py in the scripts folder of the plugin.

# api.py
from fastapi import FastAPI, Body

from modules.api.models import *
from modules.api import api
import gradio as gr

import rembg

def rembg_api(_: gr.Blocks, app: FastAPI):
    @app.post("/rembg")
    async def rembg_remove(
        input_image: str = Body("", title='rembg input image'),
        model: str = Body("u2net", title='rembg model'), 
        return_mask: bool = Body(False, title='return mask'), 
        alpha_matting: bool = Body(False, title='alpha matting'), 
        alpha_matting_foreground_threshold: int = Body(240, title='alpha matting foreground threshold'), 
        alpha_matting_background_threshold: int = Body(10, title='alpha matting background threshold'), 
        alpha_matting_erode_size: int = Body(10, title='alpha matting erode size')
    ):
        if not model or model == "None":
            return

        input_image = api.decode_base64_to_image(input_image)

        image = rembg.remove(
            input_image,
            session=rembg.new_session(model),
            only_mask=return_mask,
            alpha_matting=alpha_matting,
            alpha_matting_foreground_threshold=alpha_matting_foreground_threshold,
            alpha_matting_background_threshold=alpha_matting_background_threshold,
            alpha_matting_erode_size=alpha_matting_erode_size,
        )

        return {"image": api.encode_pil_to_base64(image).decode("utf-8")}

try:
    import modules.script_callbacks as script_callbacks

    script_callbacks.on_app_started(rembg_api)
except:
    pass
image