app.py

import gradio as gr
from PIL import Image
from vit_model_test import CustomModel  # Ensure this is the correct import for your model
from vit_Training import Custom_VIT_Model

# Initialize the model
model = CustomModel()

model_training = Custom_VIT_Model()


custom_css = """
<style>
#submit_btn {
    background-color: #4CAF50; /* Green */
    color: white;
    width:47%;
    margin-right: 3%;
}
#feedback_btn {
    background-color: #f44336; /* Red */
    color: white;
    width:47%;
    margin-left: 3%;
}
#clear_btn {
    background-color: #2196F3; /* Blue */
    color: white;
    width: 25%;
    float: left;
}
</style>
"""

# Variable to store the last prediction result
last_prediction = None

def predict(image: Image.Image):
    global last_prediction
    label, confidence = model.predict(image)
    result = "AI image" if label == 1 else "Real image"
    last_prediction = (image, label)  # Store the image and label for feedback
    return result, f"Confidence: {confidence:.2f}%"

def report_feedback():
    global last_prediction
    if last_prediction is not None:
        image, predicted_label = last_prediction
        correct_label = 1 if predicted_label == 0 else 0  # Invert the label
        print(f"Reporting feedback: predicted_label={predicted_label}, correct_label={correct_label}")  # Debugging line
        try:
            model_training.add_data(image, correct_label)  # Pass the incorrect prediction to the model
            print("Feedback recorded successfully.")  # Debugging line
            return "Feedback recorded. Thank you!"
        except Exception as e:
            print(f"Error recording feedback: {e}")  # Debugging line
            return f"Error recording feedback: {e}"
    else:
        print("No prediction available to report.")  # Debugging line
        return "No prediction available to report."

# Define the Gradio interface
with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
    gr.Markdown("### Vision Transformer Model")
    gr.Markdown("Upload an image to classify it using the Vision Transformer model.")
    
    image_input = gr.Image(type="pil", label="Upload Image")

    # Create a row for prediction and confidence outputs
    with gr.Row():
        prediction_output = gr.Textbox(label="Prediction", interactive=False)
        confidence_output = gr.Textbox(label="Confidence", interactive=False)

    # Create a row for feedback output
    with gr.Row():
        feedback_output = gr.Textbox(label="Feedback Status", interactive=False, scale=0,min_width=700)  # Add class to feedback output
    
    # Buttons
    with gr.Row():
        submit_btn = gr.Button("Submit", variant="primary", elem_id="submit_btn")
        feedback_btn = gr.Button("The model was wrong", variant="secondary", elem_id="feedback_btn")
        
    gr.Markdown("<br>")  # Adding space below buttons
    clear_btn = gr.Button("Clear", elem_id="clear_btn")  # No need for elem_classes here

    submit_btn.click(predict, inputs=image_input, outputs=[prediction_output, confidence_output])
    feedback_btn.click(report_feedback, outputs=feedback_output)

    # Clear button logic
    def clear_all():
        return None, "", "", ""  # Clear all outputs

    clear_btn.click(clear_all, outputs=[image_input, prediction_output, confidence_output, feedback_output])

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