app.py

import gradio as gr
from transformers import pipeline, AutoImageProcessor, Swinv2ForImageClassification
from torchvision import transforms
import torch

# Ensure using GPU if available
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

# Load the model and processor
image_processor = AutoImageProcessor.from_pretrained("haywoodsloan/ai-image-detector-deploy")
model = Swinv2ForImageClassification.from_pretrained("haywoodsloan/ai-image-detector-deploy")
model = model.to(device)
clf = pipeline(model=model, task="image-classification", image_processor=image_processor, device=device)

# Define class names
class_names = ['artificial', 'real']

def predict_image(img, confidence_threshold):
    # Convert the image to a PIL Image and resize it
    img = transforms.ToPILImage()(img)
    img = transforms.Resize((256, 256))(img)
    img = transforms.ToTensor()(img).unsqueeze(0).to(device)  # Add batch dimension and move to GPU
    
    # Get the prediction
    prediction = clf(img)
    
    # Process the prediction to match the class names
    result = {pred['label']: pred['score'] for pred in prediction}
    
    # Ensure the result dictionary contains both class names
    for class_name in class_names:
        if class_name not in result:
            result[class_name] = 0.0
    
    # Check if either class meets the confidence threshold
    if result['artificial'] >= confidence_threshold:
        return f"Label: artificial, Confidence: {result['artificial']:.4f}"
    elif result['real'] >= confidence_threshold:
        return f"Label: real, Confidence: {result['real']:.4f}"
    else:
        return "Uncertain Classification"

# Define the Gradio interface
image = gr.Image(label="Image to Analyze", sources=['upload'])
confidence_slider = gr.Slider(0.0, 1.0, value=0.5, step=0.01, label="Confidence Threshold")
label = gr.Label(num_top_classes=2)

gr.Interface(
    fn=predict_image,
    inputs=[image, confidence_slider],
    outputs=label,
    title="AI Generated Classification"
).launch()