app.py

# app.py (Streamlit version - Corrected)
import streamlit as st
from PIL import Image
import torch
import torchvision.transforms as transforms
import numpy as np
import subprocess

# Install pytorch_hub_examples package (Corrected)
try:
    import pytorch_hub_examples
except ModuleNotFoundError:
    subprocess.run(["git", "clone", "https://github.com/facebookresearch/pytorch_hub_examples.git"])
    subprocess.run(["cd", "pytorch_hub_examples", "&&", "python", "setup.py", "install"])  # Correct installation
    import pytorch_hub_examples

# Load the U-2-Net model
model = pytorch_hub_examples.u2net(pretrained=True)
model.eval()

# Define the transform
transform = transforms.Compose([
    transforms.Resize((320, 320)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

def remove_background(image):
    try:
        img = transform(image).unsqueeze(0)

        with torch.no_grad():
            out = model(img)
            mask = torch.sigmoid(out[0])

        mask = mask.squeeze().cpu().numpy()
        mask = (mask * 255).astype(np.uint8)
        mask = Image.fromarray(mask).convert("L")

        image = image.convert("RGBA")
        new_image = Image.new("RGBA", image.size, (255, 255, 255, 0))

        for x in range(image.width):
            for y in range(image.height):
                if mask.getpixel((x, y)) > 0:
                    new_image.putpixel((x, y), image.getpixel((x, y)))

        return new_image
    except Exception as e:
        st.error(f"Error: {e}")
        return None

st.title("Background Remover")

uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

if uploaded_file is not None:
    image = Image.open(uploaded_file).convert("RGB")  # Ensure RGB for transform
    st.image(image, caption="Uploaded Image", use_column_width=True)

    if st.button("Remove Background"):
        with st.spinner("Removing background..."):
            result_image = remove_background(image)
            if result_image:
                st.image(result_image, caption="Background Removed", use_column_width=True)