classification.py

"""
#################################
 Classification after training the Model, modules and methods in this file evaluate the performance of the trained
 model over the test dataset
 Test Data: Item (8) on https://ieee-dataport.org/open-access/flame-dataset-aerial-imagery-pile-burn-detection-using-drones-uavs 
 Tensorflow Version: 2.3.0
 GPU: Nvidia RTX 2080 Ti
 OS: Ubuntu 18.04
################################
"""
#########################################################
# import libraries

import numpy as np
import tensorflow as tf
from tensorflow.keras.models import load_model

from plotdata import plot_confusion_matrix
from config import Config_classification
from config import new_size

batch_size = Config_classification.get('batch_size')
image_size = (new_size.get('width'), new_size.get('height'))
epochs = Config_classification.get('Epochs')


#########################################################
# Function definition

def classify():
    """
    This function load the trained model from the previous task and evaluates the performance of that over the test
    data set.
    :return: None, Plot the Confusion matrix for the test data on the binary classification
    """
    test_ds = tf.keras.preprocessing.image_dataset_from_directory(
        "frames/Test", seed=1337, image_size=image_size, batch_size=batch_size, shuffle=True
    )

    model_fire = load_model('Output/Models/model_fire_resnet_not_weighted_40_no_metric_simple')

    _ = model_fire.evaluate(test_ds, batch_size=batch_size)

    best_model_fire = load_model('Output/Models/h5model/keras/save_at_25.h5')
    results_eval = best_model_fire.evaluate(test_ds, batch_size=batch_size)

    for name, value in zip(model_fire.metrics_names, results_eval):
        print(name, ': ', value)
    print()

    cm = np.array([[results_eval[1], results_eval[4]], [results_eval[2], results_eval[3]]])
    cm_plot_labels = ['Fire', 'No Fire']
    plot_confusion_matrix(cm=cm, classes=cm_plot_labels, title='Confusion Matrix')

    model_file = 'Output/Models/h5model/Keras_not_weighted_40_no_metric_simple/save_at_%d.h5' % 37
    model_fire = load_model(model_file)
    test_fire_ds = tf.keras.preprocessing.image_dataset_from_directory(
        "frames/confusion_test/Fire_test", seed=1337, image_size=image_size, batch_size=batch_size, shuffle=True)
    test_no_fire_ds = tf.keras.preprocessing.image_dataset_from_directory(
        "frames/confusion_test/No_Fire_test", seed=1337, image_size=image_size, batch_size=batch_size, shuffle=True)
    fire_eval = model_fire.evaluate(test_fire_ds)
    no_fire_eval = model_fire.evaluate(test_no_fire_ds)
    true_fire = len(tf.io.gfile.listdir("frames/confusion_test/Fire_test/Fire"))
    true_no_fire = len(tf.io.gfile.listdir("frames/confusion_test/No_Fire_test/No_Fire"))
    tp = fire_eval[1] * true_fire
    fp = (1 - fire_eval[1]) * true_fire
    tn = (1 - no_fire_eval[1]) * true_no_fire
    fn = no_fire_eval[1] * true_no_fire
    cm = np.array([[tp, fn], [fp, tn]], dtype=int)
    plot_confusion_matrix(cm=cm, classes=cm_plot_labels, title='Confusion Matrix')