app.py

import gradio as gr
import tensorflow as tf
from PIL import Image
import numpy as np
from PIL import Image

# Load the model outside of the prediction function to avoid reloading it on each request
MODEL = tf.keras.models.load_model('my_final_model.h5')
CLASS_NAMES = ['Potato_Early_blight', 'Potato_Late_blight', 'Potato_healthy']



def classify_image(image_input):
    # Convert the NumPy array to a PIL Image
    image_input = Image.fromarray(image_input.astype('uint8'), 'RGB')

    # Resize the image
    image_processed = image_input.resize((256, 256))

    # Convert the PIL Image back to NumPy array and preprocess it
    image_processed = np.array(image_processed) / 255.0
    image_processed = image_processed.reshape((1, 256, 256, 3))

    # Make a prediction
    predictions = MODEL.predict(image_processed)
    prediction_result = {CLASS_NAMES[i]: float(predictions[0][i]) for i in range(len(CLASS_NAMES))}

    # Prepare the markdown information
    blight_info = """
    ### Late Blight in Potatoes
    **Pathogen:** Fungus
    **Hosts:** Potatoes
    ...
    For more information and updates on potato blight strains, visit [EuroBlight](www.euroblight.net).
    """

    return prediction_result, blight_info


def main():
    # Define Gradio interface components
    image_input = gr.Image(label="Upload a Potato Leaf Image")
    outputs = gr.Label(num_top_classes=3, label="Predictions")
    blight_info_markdown = gr.Markdown()

    # Create Gradio interface
    interface = gr.Interface(
        fn=classify_image,
        inputs=image_input,
        outputs=[outputs, blight_info_markdown],
        title="Potato Leaf Disease Detection",
        description="Upload an image of a potato leaf, and the model will predict if it is healthy or has early blight or late blight."
    )

     # Launch the interface
    interface.launch(share=True)

if __name__ == "__main__":
    main()