test_removal.py

import numpy as np
from PIL import Image
from transformers import pipeline
import matplotlib.pyplot as plt

# List of models to test
MODELS = {
    "BRIA": "BRIA-AI/bria-rmbg",
    "INSPyReNet": "mattmdjaga/INSPyReNet",
    "U2Net": "silks-road/u2net",
    "U2Net-Human": "mattmdjaga/u2net-human-seg",
    "ISNet-General": "xuebinqin/ISNet-general-use",
    "ISNet-Anime": "skytnt/anime-seg"
}

def test_removal(image_path):
    # Load image
    original = Image.open(image_path).convert("RGB")
    img_np = np.array(original)
    
    # Create figure to display results
    plt.figure(figsize=(15, 8))
    plt.subplot(2, 4, 1)
    plt.imshow(original)
    plt.title("Original")
    plt.axis('off')
    
    # Test each model
    for i, (name, repo) in enumerate(MODELS.items(), start=2):
        try:
            print(f"Testing {name}...")
            pipe = pipeline("image-segmentation", repo)
            result = pipe(img_np)
            mask = result[0]['mask'] if isinstance(result, list) else result['mask']
            
            # Apply mask
            background = Image.new('RGB', original.size, (255, 0, 0))  # Red background for visibility
            result_img = Image.composite(original, background, Image.fromarray(mask))
            
            plt.subplot(2, 4, i)
            plt.imshow(result_img)
            plt.title(name)
            plt.axis('off')
            print(f"{name} worked!")
        except Exception as e:
            print(f"{name} failed: {str(e)}")
            plt.subplot(2, 4, i)
            plt.text(0.5, 0.5, f"{name}\nFailed", ha='center')
            plt.axis('off')
    
    plt.tight_layout()
    plt.show()

# Test with your image
test_removal("your_image.jpg")  # Change this to your image path