app.py

import os
import gradio as gr
import tensorflow as tf
import numpy as np

os.environ['CUDA_VISIBLE_DEVICES'] = '-1'

# Loading saved model
model = tf.keras.models.load_model('plant_disease_classifier.h5')

def predict(input_image):
    try:
        # Preprocessing
        input_image = tf.convert_to_tensor(input_image)
        input_image = tf.image.resize(input_image, [256, 256])
        input_image = tf.expand_dims(input_image, 0) / 255.0

     
        predictions = model.predict(input_image)
        labels = ['Healthy', 'Powdery', 'Rust']

       
        class_idx = np.argmax(predictions)
        class_label = labels[class_idx]
        confidence = np.round(predictions[0][class_idx] * 100, 3)

        return f"Predicted Class: {class_label}.  Confidence Score: {confidence}%"

    except Exception as e:
        return f"An error occurred: {e}"

examples = ["Healthy.png", "Powdery.png", "Rust.png"]

iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(),
    outputs="text",
    title="🌿 Plant Disease Detection",
    description='<br> This is a specialized Image Classification model engineered to identify the health status of plants, specifically detecting conditions of Powdery Mildew or Rust. <br> \
    This model is based on a Convolutional Neural Network that I have trained, evaluated, and validated on my Kaggle Notebook: <a href="https://www.kaggle.com/code/lusfernandotorres/convolutional-neural-network-from-scratch">🧠 Convolutional Neural Network From Scratch</a>. <br> \
    <br> Upload a photo of a plant to see how the model classifies its status!',
    examples=examples
)

iface.launch(share=True)