app.py

import os
import zipfile
import gradio as gr
import tensorflow as tf
import numpy as np
from tensorflow.keras.preprocessing import image
from tensorflow.keras.layers import TFSMLayer
from PIL import Image

# Define paths
zip_path = "best_model_saved.zip"
extract_path = "best_model_saved"

# Check if extracted folder exists, if not, extract it
if not os.path.exists(extract_path):
    with zipfile.ZipFile(zip_path, 'r') as zip_ref:
        zip_ref.extractall(".")

# Load the model using TFSMLayer
best_model = TFSMLayer(extract_path, call_endpoint="serving_default")
print("✅ Model loaded successfully!")

# Define class mapping (index -> class label)
class_mapping = {0: "Early_Blight", 1: "Healthy", 2: "Late_Blight"}

def classify_leaf(input_image):
    """
    Takes an input image (PIL format from Gradio), preprocesses it,
    and returns the predicted class with probabilities.
    """
    # Resize and normalize the image
    img = input_image.resize((250, 250))
    img_array = image.img_to_array(img) / 255.0
    img_array = np.expand_dims(img_array, axis=0)  # shape (1, 250, 250, 3)

    # Make a prediction
    pred_dict = best_model(img_array)  # Get dictionary output

    # Extract the correct tensor output from the dictionary
    output_key = list(pred_dict.keys())[0]  # Get first output key
    preds = pred_dict[output_key]  # Extract tensor

    # Convert tensor to numpy array
    probs = preds.numpy()[0]  # Convert to NumPy and extract first batch

    # Format probabilities as a string
    prob_str = ", ".join([f"{class_mapping[i]}: {probs[i]:.2f}" for i in range(len(class_mapping))])

    # Determine predicted class
    pred_idx = np.argmax(probs)
    pred_class = class_mapping[pred_idx]

    # Return formatted result
    return f"Predicted Class: {pred_class}\nProbabilities: {prob_str}"

# Build the Gradio interface
iface = gr.Interface(
    fn=classify_leaf,
    inputs=gr.Image(type="pil"),   # Expecting an uploaded image
    outputs="text",                # Output text with the predicted class and probabilities
    title="Potato Leaf Disease Classifier",
    description="Upload an image of a potato leaf. The model will predict if it's Early_Blight, Healthy, or Late_Blight."
)

# Launch the app
if __name__ == "__main__":
    iface.launch()