app.py

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

# 모델 로드 및 CPU로 설정
birefnet = AutoModelForImageSegmentation.from_pretrained(
    "ZhengPeng7/BiRefNet", trust_remote_code=True
)
birefnet.to("cpu")  # GPU -> CPU로 변경

# 이미지 전처리
transform_image = transforms.Compose(
    [
        transforms.Resize((1024, 1024)),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
    ]
)

def process(image):
    image_size = image.size
    input_images = transform_image(image).unsqueeze(0).to("cpu")  # CPU로 변경
    # 예측 수행
    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)
    image.putalpha(mask)
    return image

def fn(image):
    im = load_img(image, output_type="pil")
    im = im.convert("RGB")
    origin = im.copy()
    processed_image = process(im)
    
    # JPG로 변환하여 저장
    jpg_image = origin.copy()
    jpg_image = jpg_image.convert("RGB")
    jpg_path = "output.jpg"
    jpg_image.save(jpg_path, format="JPEG")
    
    return [processed_image], jpg_path  # ImageSlider는 리스트를 기대함

def process_file(f):
    name_path = f.rsplit(".", 1)[0] + ".png"
    im = load_img(f, output_type="pil")
    im = im.convert("RGB")
    transparent = process(im)
    transparent.save(name_path)
    return name_path

# Gradio 컴포넌트 정의
slider1 = ImageSlider(label="Processed Image", type="pil")
image_upload = gr.Image(label="Upload an image")
output_download = gr.File(label="Download JPG File")

# 새로운 샘플 이미지 추가 (app.py와 동일한 폴더에 위치해야 함)
sample_images = ["1.png", "2.jpg", "3.png"]

# Gradio 인터페이스 설정
tab1 = gr.Interface(
    fn=fn, 
    inputs=image_upload, 
    outputs=[slider1, output_download], 
    examples=sample_images, 
    api_name="image"
)

demo = gr.Interface(
    tab1, 
    title="Background Removal Tool", 
    description="이미지를 업로드하면 배경이 제거된 이미지를 확인하고 JPG 파일로 다운로드할 수 있습니다."
)

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