|
from PIL import Image |
|
import requests |
|
import os |
|
|
|
from transformers import CLIPSegProcessor, CLIPSegForImageSegmentation |
|
from diffusers import DiffusionPipeline |
|
import torch |
|
from torch import autocast |
|
|
|
import gradio as gr |
|
|
|
auth_token = os.environ.get("API_TOKEN") or True |
|
|
|
url = "https://github.com/timojl/clipseg/blob/master/example_image.jpg?raw=true" |
|
image = Image.open(requests.get(url, stream=True).raw) |
|
|
|
processor = CLIPSegProcessor.from_pretrained("CIDAS/clipseg-rd64-refined") |
|
model = CLIPSegForImageSegmentation.from_pretrained("CIDAS/clipseg-rd64-refined") |
|
|
|
pipe = DiffusionPipeline.from_pretrained( |
|
"runwayml/stable-diffusion-inpainting", |
|
custom_pipeline="text_inpainting", |
|
segmentation_model=model, |
|
segmentation_processor=processor, |
|
use_auth_token=auth_token, |
|
) |
|
|
|
device = "cuda" if torch.cuda.is_available() else "cpu" |
|
pipe = pipe.to(device) |
|
|
|
def pad_image(image): |
|
w, h = image.size |
|
if w == h: |
|
return image |
|
elif w > h: |
|
new_image = Image.new(image.mode, (w, w), (0, 0, 0)) |
|
new_image.paste(image, (0, (w - h) // 2)) |
|
return new_image |
|
else: |
|
new_image = Image.new(image.mode, (h, h), (0, 0, 0)) |
|
new_image.paste(image, ((h - w) // 2, 0)) |
|
return new_image |
|
|
|
|
|
def process_image(image, text, prompt): |
|
image = pad_image(image) |
|
image = image.resize((512, 512)) |
|
with autocast("cuda"): |
|
inpainted_image = pipe(image=image, text=text, prompt=prompt).images[0] |
|
return inpainted_image |
|
|
|
|
|
title = "Interactive demo: Text-based inpainting with CLIPSeg x Stable Diffusion" |
|
description = "Demo for using CLIPSeg, a CLIP-based model for zero- and one-shot image segmentation. This model can be used to segment things in an image based on text. This way, one can use it to provide a binary mask for Stable Diffusion, which the latter needs to inpaint. To use it, simply upload an image and add a text to mask as well as a text which indicates what to replace, or use one of the examples below and click 'submit'. Results will show up in a few seconds." |
|
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2112.10003'>CLIPSeg: Image Segmentation Using Text and Image Prompts</a> | <a href='https://huggingface.co/docs/transformers/main/en/model_doc/clipseg'>HuggingFace docs</a></p>" |
|
|
|
examples = [["example_image.png", "a glass", "a cup"]] |
|
|
|
interface = gr.Interface(fn=process_image, |
|
inputs=[gr.Image(type="pil"), gr.Textbox(label="What's the thing you want to replace?"), gr.Textbox(label="What do you want as replacement?")], |
|
outputs=gr.Image(type="pil"), |
|
title=title, |
|
description=description, |
|
article=article, |
|
examples=examples) |
|
|
|
interface.launch(debug=True) |
