# Breast Cancer Detection Model Analysis Validation

import streamlit as st
import numpy as np
import tensorflow as tf 

# Load the trained model CNN_model.h5
model = tf.keras.models.load_model('CNN_model.h5')

# Load the trained model X_test.npy
X_test = np.load('X_test.npy')

# Load the trained model y_test.npy 
y_test = np.load('y_test.npy')

# Populate the X_test shape
st.write('The shape of X_test = ', X_test.shape)

# Populate the y_test shape
st.write('The shape of y_test = ', y_test.shape)

model.summary()

# Populate the model summary
st.write('The model summary =', model)

# The output of y_test equals (55505, 2) 2-Dimensional Array [Reduce the Dimensionality into 1D]
# Ensure y_test is a 1D array (Although, the inference drawn from y_test data is that it is a 2D array)
if len(y_test.shape) > 1:
    st.write('The y_test.shape is more than 1 Dimension.')
    st.write('Reducing Dimensionality in y_test.shape array...')
    y_test = y_test.flatten() # Flatten the array if it is not 1D
    st.write('The y_test.shape after Dimensionality Reduction (Flattening): ', y_test.shape)
    st.write('Dimensionality Reduction on y_test complete!') 

# Define a mapping of class indices to human-readable labels
class_labels = {
    0: 'Non-Cancerous',
    1: 'Cancerous',
}

# Define a function make predictions
def predict_diagnosis(image_data, image_index):
    # Preprocess the image data (e.g., reshape, normalize)
    if len(image_data.shape) == 3:
        st.write('Implementation fails with: # image_data = image_data.reshape(1, image_data.shape[0], image_data.shape[1], image_data.shape[2])')
        st.write()
        st.write('Reshaping 3-Dimensional array into 4-Dimensional array not possible')
        st.write()
        st.write('Reshape an array of size 56250000 into a shape of (1,1,50,50)')
        st.write()
        st.write('Not possible because the total number of elements in the array (56250000)')
        st.write()
        st.write("Does not match the product of the new shapes dimensions (1150*50 = 5000)")
        st.write()
        st.write('Adding a new dimension obverse np.expand_dims not reshape')
        st.write()
        image_data = np.expand_dims(image_data, axis=0)

    image_data = image_data.astype('float32') / 255 

    # Make a prediction 
    prediction = model.predict(image_data)

    # Get the predicted label index 
    predicted_label_index = np.argmax(prediction) # Returns Highest Probability
    predicted_label = class_labels[predicted_label_index] 

    # Convert the true label to an integer
    true_label_index = int(y_test[image_index]) 
    true_label = class_labels[true_label_index] 

    return predicted_label, true_label 

# Select an image index from the X_test dataset: 89
image_index = 89
image_data = X_test[image_index]

# Predict the diagnosis
predicted_label, true_label = predict_diagnosis(image_data, image_index=image_index) 

st.write('Predicted Diagnosis:', predicted_label)
st.write('True Diagnosis:', true_label)