Spaces:
Running
Running
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from PIL import Image
|
| 3 |
+
import requests
|
| 4 |
+
from io import BytesIO
|
| 5 |
+
import torch
|
| 6 |
+
from torchvision import transforms
|
| 7 |
+
from diffusers import AutoencoderKL, LCMScheduler
|
| 8 |
+
from pipeline_controlnet_sd_xl import StableDiffusionXLControlNetPipeline
|
| 9 |
+
from controlnet import ControlNetModel
|
| 10 |
+
|
| 11 |
+
# Define helper functions
|
| 12 |
+
def download_image(url):
|
| 13 |
+
response = requests.get(url)
|
| 14 |
+
return Image.open(BytesIO(response.content)).convert("RGB")
|
| 15 |
+
|
| 16 |
+
def load_model():
|
| 17 |
+
# Load model components
|
| 18 |
+
controlnet = ControlNetModel().from_pretrained("briaai/DEV-ControlNetInpaintingFast", torch_dtype=torch.float16)
|
| 19 |
+
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
|
| 20 |
+
pipe = StableDiffusionXLControlNetPipeline.from_pretrained("briaai/BRIA-2.3", controlnet=controlnet.to(dtype=torch.float16), torch_dtype=torch.float16, vae=vae)
|
| 21 |
+
pipe.to('cuda')
|
| 22 |
+
return pipe
|
| 23 |
+
|
| 24 |
+
pipe = load_model()
|
| 25 |
+
|
| 26 |
+
# Define the inpainting function
|
| 27 |
+
def inpaint(image, mask):
|
| 28 |
+
# Process image and mask
|
| 29 |
+
image = image.resize((1024, 1024)).convert("RGB")
|
| 30 |
+
mask = mask.resize((1024, 1024)).convert("L")
|
| 31 |
+
|
| 32 |
+
# Transform to tensor
|
| 33 |
+
image_transform = transforms.ToTensor()
|
| 34 |
+
image_tensor = image_transform(image).unsqueeze(0).to('cuda')
|
| 35 |
+
mask_tensor = image_transform(mask).unsqueeze(0).to('cuda')
|
| 36 |
+
mask_tensor = (mask_tensor > 0.5).float() # binarize mask
|
| 37 |
+
|
| 38 |
+
# Generate image
|
| 39 |
+
with torch.no_grad():
|
| 40 |
+
result = pipe(prompt="A park bench", init_image=image_tensor, mask_image=mask_tensor, num_inference_steps=50).images[0]
|
| 41 |
+
|
| 42 |
+
return transforms.ToPILImage()(result.squeeze(0))
|
| 43 |
+
|
| 44 |
+
# Define the interface
|
| 45 |
+
interface = gr.Interface(fn=inpaint,
|
| 46 |
+
inputs=[gr.inputs.Image(type="pil", label="Original Image"), gr.inputs.Image(type="pil", label="Mask Image")],
|
| 47 |
+
outputs=gr.outputs.Image(type="pil", label="Inpainted Image"),
|
| 48 |
+
title="Stable Diffusion XL ControlNet Inpainting",
|
| 49 |
+
description="Upload an image and its corresponding mask to inpaint the specified area.")
|
| 50 |
+
|
| 51 |
+
if __name__ == "__main__":
|
| 52 |
+
interface.launch()
|