app.py

import numpy as np
import gradio as gr
import tensorflow as tf  # version 2.13.0
from keras.models import load_model
import cv2
import json
import os

def analyse(img, plant_type):
    # Load label_disease.json
    with open('data/label_disease.json', 'r') as f:
        label_disease = json.load(f)

    # Load plant_label_disease.json
    with open('data/plant_label_disease.json', 'r') as f:
        plant_label_disease = json.load(f)

    HEIGHT = 256
    WIDTH = 256
    modelArchitecturePath = 'model/model_architecture.h5'
    modelWeightsPath = 'model/model_weights.h5'

    # Load the model
    dnn_model = load_model(modelArchitecturePath, compile=False)
    dnn_model.load_weights(modelWeightsPath)

    # Preprocess the image
    process_img = cv2.resize(img, (HEIGHT, WIDTH), interpolation=cv2.INTER_LINEAR)
    process_img = process_img / 255.0
    process_img = np.expand_dims(process_img, axis=0)

    # Predict using the model
    y_pred = dnn_model.predict(process_img)
    y_pred = y_pred[0]

    # Identify plant-specific predictions
    plant_label_ids = plant_label_disease[plant_type.lower()]
    plant_predicted_id = plant_label_ids[0]
    for disease in plant_label_ids:
        if y_pred[disease] > y_pred[plant_predicted_id]:
            plant_predicted_id = disease

    # Determine overall prediction
    overall_predicted_id = int(np.argmax(y_pred))
    overall_predicted_name = label_disease[str(overall_predicted_id)]
    overall_predicted_confidence = float(y_pred[overall_predicted_id])

    # Determine plant-specific prediction
    plant_predicted_name = label_disease[str(plant_predicted_id)]
    plant_predicted_confidence = float(y_pred[plant_predicted_id])

    # Determine health status
    is_plant_specific_healthy = "healthy" in plant_predicted_name.lower()
    is_overall_healthy = "healthy" in overall_predicted_name.lower()

    # Return results as a JSON object
    result = {
        "plant_specific_prediction_id": plant_predicted_id,
        "plant_specific_prediction_name": plant_predicted_name,
        "plant_specific_confidence": plant_predicted_confidence,
        "is_plant_specific_healthy": is_plant_specific_healthy,
        "overall_prediction_id": overall_predicted_id,
        "overall_prediction_name": overall_predicted_name,
        "overall_confidence": overall_predicted_confidence,
        "is_overall_healthy": is_overall_healthy
    }

    return result

# Create Gradio interface
input_image = gr.Image(type="numpy")
plant_type = gr.Radio(["Apple", "Blueberry", "Cherry", "Corn", "Grape", "Orange", "Peach",
                     "Pepper", "Potato", "Raspberry", "Soybean", "Squash", "Strawberry", "Tomato"])

demo = gr.Interface(
    fn=analyse,
    inputs=[input_image, plant_type],
    outputs=gr.JSON(),
    title="Plant Disease Detection",
    description="Upload an image of a plant leaf or use one of the example images below. Select the type of plant and the model will analyze it for diseases."
)

# Add examples component
examples = gr.Examples(
    examples=[
        os.path.join("examples", img_name) for img_name in sorted(os.listdir("examples"))
    ],
    inputs=[input_image],
    label="Examples",
    cache_examples=False,
    examples_per_page=16
)

# Launch the interface
demo.launch(share=True, show_error=True)