app.py

import gradio as gr
from gradio_imageslider import ImageSlider
import spaces
from transformers import AutoModelForImageSegmentation
import torch
from torchvision import transforms
from PIL import Image

torch.set_float32_matmul_precision(["high", "highest"][0])

birefnet = AutoModelForImageSegmentation.from_pretrained(
    "ZhengPeng7/BiRefNet", trust_remote_code=True
)
birefnet.to("cuda")
transform_image = transforms.Compose(
    [
        transforms.Resize((1024, 1024)),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
    ]
)


@spaces.GPU
def fn(image):
    if image is None or len(image) == 0:
        return image, None  # 원본 이미지도 반환
    im = Image.open(image).convert("RGB")
    image_size = im.size
    origin = im.copy()
    input_images = transform_image(im).unsqueeze(0).to("cuda")
    # Prediction
    with torch.no_grad():
        preds = birefnet(input_images)[-1].sigmoid().cpu()
    pred = preds[0].squeeze()
    pred_pil = transforms.ToPILImage()(pred)
    mask = pred_pil.resize(image_size)
    im.putalpha(mask)
    return im, origin  # 변환된 이미지와 원본 이미지 반환


def save_image(image):
    if image is not None:
        image.save("output.png")
        return "output.png"
    return None


with gr.Blocks() as demo:
    image = gr.Image(label="Upload an image")
    text = gr.Textbox(label="Paste an image URL")
    download_button = gr.Button("Download Image")
    output_file = gr.File()
    slider1 = ImageSlider(label="birefnet", type="pil")
    slider2 = ImageSlider(label="birefnet", type="pil")

    # 스페이스에 있는 예제 이미지 파일 경로
    example_image1 = "example_images/example1.jpg"
    example_image2 = "example_images/example2.jpg"
    example_image3 = "example_images/example3.jpg"
    
    with gr.Tab("Image Upload"):
        tab1 = gr.Interface(
            fn, inputs=image, outputs=[slider1, output_file], 
            examples=[example_image1, example_image2, example_image3], api_name="image"
        )

    with gr.Tab("Image URL"):
        tab2 = gr.Interface(
            fn, inputs=text, outputs=[slider2, output_file], 
            examples=[example_image1, example_image2, example_image3], api_name="text"
        )

    def process_download(image):
        return save_image(image[0])

    download_button.click(process_download, inputs=slider1, outputs=output_file)

if __name__ == "__main__":
    demo.launch()