app.py

import gradio as gr
import numpy as np
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
from PIL import Image

# Load your model
model = load_model("Model.hdf5")

# Class mapping (from your screenshot)
class_indices = {
    'Apple___Apple_scab': 0,
    'Apple___Black_rot': 1,
    'Apple___Cedar_apple_rust': 2,
    'Apple___healthy': 3,
    'Blueberry___healthy': 4,
    'Cherry_(including_sour)___Powdery_mildew': 5,
    'Cherry_(including_sour)___healthy': 6,
    'Corn_(maize)___Cercospora_leaf_spot Gray_leaf_spot': 7,
    'Corn_(maize)___Common_rust': 8,
    'Corn_(maize)___Northern_Leaf_Blight': 9,
    'Corn_(maize)___healthy': 10,
    'Grape___Black_rot': 11,
    'Grape___Esca_(Black_Measles)': 12,
    'Grape___Leaf_blight_(Isariopsis_Leaf_Spot)': 13,
    'Grape___healthy': 14,
    'Orange___Haunglongbing_(Citrus_greening)': 15,
    'Peach___Bacterial_spot': 16,
    'Peach___healthy': 17,
    'Pepper,_bell___Bacterial_spot': 18,
    'Pepper,_bell___healthy': 19,
    'Potato___Early_blight': 20,
    'Potato___Late_blight': 21,
    'Potato___healthy': 22,
    'Raspberry___healthy': 23,
    'Soybean___healthy': 24,
    'Squash___Powdery_mildew': 25,
    'Strawberry___Leaf_scorch': 26,
    'Strawberry___healthy': 27,
    'Tomato___Bacterial_spot': 28,
    'Tomato___Early_blight': 29,
    'Tomato___Late_blight': 30,
    'Tomato___Leaf_Mold': 31,
    'Tomato___Septoria_leaf_spot': 32,
    'Tomato___Spider_mites Two-spotted_spider_mite': 33,
    'Tomato___Target_Spot': 34,
    'Tomato___Tomato_Yellow_Leaf_Curl_Virus': 35,
    'Tomato___Tomato_mosaic_virus': 36,
    'Tomato___healthy': 37
}

# Reverse mapping
idx_to_class = {v: k for k, v in class_indices.items()}

# Prediction function
def predict(img):
    img = img.resize((224, 224))  # adjust if your model uses different input size
    img_array = image.img_to_array(img) / 255.0
    img_array = np.expand_dims(img_array, axis=0)

    predictions = model.predict(img_array)
    predicted_class = np.argmax(predictions[0])
    confidence = np.max(predictions[0]) * 100

    return f"{idx_to_class[predicted_class]} ({confidence:.2f}% confidence)"

# Gradio interface
demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil", label="Upload a leaf image"),
    outputs=gr.Textbox(label="Prediction"),
    title="Crop Disease Prediction",
    description="Upload an image of a crop leaf to predict its disease."
)

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