app.py

import gradio as gr
import numpy as np
import cv2
from tensorflow.keras.models import load_model
from tensorflow.keras.layers import InputLayer

# Custom object scope to handle InputLayer configuration
custom_objects = {'InputLayer': InputLayer}

# Load model
def load_ocular_model():
    return load_model('odir_cnn_model.h5', custom_objects=custom_objects)

model = load_ocular_model()

# Define the labels
LABELS = ['Normal (N)', 'Diabetes (D)', 'Glaucoma (G)', 'Cataract (C)', 
          'Age related Macular Degeneration (A)', 'Hypertension (H)', 
          'Pathological Myopia (M)', 'Other diseases/abnormalities (O)']

# Preprocess the image
def preprocess_image(image, img_size=128):  # Change img_size to 128
    img = cv2.resize(image, (img_size, img_size))
    img = img / 255.0  # Normalize
    img = np.expand_dims(img, axis=0)  # Add batch dimension
    return img

# Predict diseases based on left and right images
def predict_diseases(left_image, right_image):
    left_img = preprocess_image(left_image)
    right_img = preprocess_image(right_image)
    
    left_predictions = model.predict(left_img)
    right_predictions = model.predict(right_img)
    
    combined_predictions = (left_predictions + right_predictions) / 2
    
    pred_labels = {label: float(pred) for label, pred in zip(LABELS, combined_predictions[0])}
    
    return pred_labels

# Define the Gradio interface
def predict_from_images(left_image, right_image):
    left_image = cv2.cvtColor(np.array(left_image), cv2.COLOR_RGB2BGR)  # Convert Gradio's image to OpenCV format
    right_image = cv2.cvtColor(np.array(right_image), cv2.COLOR_RGB2BGR)  # Convert Gradio's image to OpenCV format
    predictions = predict_diseases(left_image, right_image)
    
    return predictions

# Gradio interface
input_images = [gr.Image(type="pil", label="Left Eye Image"), gr.Image(type="pil", label="Right Eye Image")]
output_labels = gr.Label(num_top_classes=len(LABELS), label="Ocular Disease Predictions")

gr_interface = gr.Interface(fn=predict_from_images, 
                            inputs=input_images, 
                            outputs=output_labels,
                            title="Ocular Disease Prediction",
                            description="Upload left and right eye images to predict ocular diseases.",
                            live=False)

# Launch the Gradio app
gr_interface.launch()