app.py

import streamlit as st 
import numpy as np 
import pickle 
from tensorflow.keras.models import load_model 

# Load the model
@st.cache_resource() 
def load_h5_model(model_path):
    return load_model(model_path) 

# Load the data 
@st.cache_resource()
def load_npy_data(data_path):
    return np.load(data_path) 

# Load the pickle file
@st.cache_resource()
def load_pkl_file(pkl_path):
    with open(pkl_path, 'rb') as f:
        return pickle.load(f) 

# Streamlit app
def main():
    st.title("Breast Cancer Detection Model Evaluation")

    # Load the model
    model = load_h5_model('CNN_model.h5')

    # Load the data 
    X_test = load_npy_data('X_test.npy') 
    y_test = load_npy_data('y_test.npy') 

    # Load the accuracy from accuracy_epoch_5000.pkl 
    loaded_correct_predictions, loaded_correct_indices = load_pkl_file('accuracy_epoch_5000.pkl')

    # Load the history from history.pkl 
    history = load_pkl_file('history.pkl')

    # Display the shapes of X_test and y_test 
    st.write("X_test shape:", X_test.shape)
    st.write("y_test shape:", y_test.shape) 

    # Display the model summary (utilize st.text not st.write)
    st.write("CNN_model.h5 Summary")
    st.text(model.summary())

    # Display the loaded accuracy results 
    st.write("accuracy_epoch_5000 results:")
    st.write("Loaded Correct Predictions:", loaded_correct_predictions)
    st.write("Loaded Correct Indices:", loaded_correct_indices) 

    # Display the model accuracy from history
    # (55005 - 12171) / 55005 = 0.78 % 
    # Adding rest of epochs from 55005 images
    # Then, drawing inference on how close 
    # in proximity the Accuracy and History pickle file are
    # The history pickle file is 
    st.write("Breast Cancer Detection Model Accuracy deriving from history:")
    if 'accuracy' in history:
        st.line_chart(history['accuracy']) 
    else:
        st.write("No accuracy data available in history.")

if __name__ == "__main__":
    main()