in_painting_with_stable_diffusion_using_diffusers.ipynb

Kernel: Python 3

In-painting pipeline for Stable Diffusion using 🧨 Diffusers

This notebook shows how to do text-guided in-painting with Stable Diffusion model using 🤗 Hugging Face 🧨 Diffusers library.

For a general introduction to the Stable Diffusion model please refer to this colab.

!pip install -qq -U diffusers==0.11.1 transformers ftfy gradio accelerate

To use private and gated models on 🤗 Hugging Face Hub, login is required. If you are only using a public checkpoint (such as runwayml/stable-diffusion-inpainting in this notebook), you can skip this step.

from huggingface_hub import notebook_login

notebook_login()

Login successful
Your token has been saved to /root/.huggingface/token

import inspect
from typing import List, Optional, Union

import numpy as np
import torch

import PIL
import gradio as gr
from diffusers import StableDiffusionInpaintPipeline

device = "cuda"
model_path = "runwayml/stable-diffusion-inpainting"

pipe = StableDiffusionInpaintPipeline.from_pretrained(
    model_path,
    torch_dtype=torch.float16,
).to(device)

Fetching 15 files:   0%|          | 0/15 [00:00<?, ?it/s]

import requests
from io import BytesIO

def image_grid(imgs, rows, cols):
    assert len(imgs) == rows*cols

    w, h = imgs[0].size
    grid = PIL.Image.new('RGB', size=(cols*w, rows*h))
    grid_w, grid_h = grid.size
    
    for i, img in enumerate(imgs):
        grid.paste(img, box=(i%cols*w, i//cols*h))
    return grid


def download_image(url):
    response = requests.get(url)
    return PIL.Image.open(BytesIO(response.content)).convert("RGB")

img_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png"
mask_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo_mask.png"

image = download_image(img_url).resize((512, 512))
image

mask_image = download_image(mask_url).resize((512, 512))
mask_image

prompt = "a mecha robot sitting on a bench"

guidance_scale=7.5
num_samples = 3
generator = torch.Generator(device="cuda").manual_seed(0) # change the seed to get different results

images = pipe(
    prompt=prompt,
    image=image,
    mask_image=mask_image,
    guidance_scale=guidance_scale,
    generator=generator,
    num_images_per_prompt=num_samples,
).images

  0%|          | 0/50 [00:00<?, ?it/s]

# insert initial image in the list so we can compare side by side
images.insert(0, image)

image_grid(images, 1, num_samples + 1)

Gradio Demo

def predict(dict, prompt):
  image =  dict['image'].convert("RGB").resize((512, 512))
  mask_image = dict['mask'].convert("RGB").resize((512, 512))
  images = pipe(prompt=prompt, image=image, mask_image=mask_image).images
  return(images[0])

gr.Interface(
    predict,
    title = 'Stable Diffusion In-Painting',
    inputs=[
        gr.Image(source = 'upload', tool = 'sketch', type = 'pil'),
        gr.Textbox(label = 'prompt')
    ],
    outputs = [
        gr.Image()
        ]
).launch(debug=True)

Colab notebook detected. This cell will run indefinitely so that you can see errors and logs. To turn off, set debug=False in launch().
Running on public URL: https://e52f060882d60b09.gradio.app

This share link expires in 72 hours. For free permanent hosting and GPU upgrades (NEW!), check out Spaces: https://huggingface.co/spaces

  0%|          | 0/50 [00:00<?, ?it/s]

  0%|          | 0/50 [00:00<?, ?it/s]

Keyboard interruption in main thread... closing server.

(<gradio.routes.App at 0x7eff58b4ef90>,
 'http://127.0.0.1:7860/',
 'https://e52f060882d60b09.gradio.app')