Python/computer_vision/cnn_classification.py

"""
Convolutional Neural Network

Objective : To train a CNN model detect if TB is present in Lung X-ray or not.

Resources CNN Theory :
    https://en.wikipedia.org/wiki/Convolutional_neural_network
Resources Tensorflow : https://www.tensorflow.org/tutorials/images/cnn

Download dataset from :
https://lhncbc.nlm.nih.gov/LHC-publications/pubs/TuberculosisChestXrayImageDataSets.html

1. Download the dataset folder and create two folder training set and test set
in the parent dataset folder
2. Move 30-40 image from both TB positive and TB Negative folder
in the test set folder
3. The labels of the images will be extracted from the folder name
the image is present in.

"""

# Part 1 - Building the CNN

import numpy as np

# Importing the Keras libraries and packages
import tensorflow as tf
from tensorflow.keras import layers, models

if __name__ == "__main__":
    # Initialising the CNN
    # (Sequential- Building the model layer by layer)
    classifier = models.Sequential()

    # Step 1 - Convolution
    # Here 64,64 is the length & breadth of dataset images and 3 is for the RGB channel
    # (3,3) is the kernel size (filter matrix)
    classifier.add(
        layers.Conv2D(32, (3, 3), input_shape=(64, 64, 3), activation="relu")
    )

    # Step 2 - Pooling
    classifier.add(layers.MaxPooling2D(pool_size=(2, 2)))

    # Adding a second convolutional layer
    classifier.add(layers.Conv2D(32, (3, 3), activation="relu"))
    classifier.add(layers.MaxPooling2D(pool_size=(2, 2)))

    # Step 3 - Flattening
    classifier.add(layers.Flatten())

    # Step 4 - Full connection
    classifier.add(layers.Dense(units=128, activation="relu"))
    classifier.add(layers.Dense(units=1, activation="sigmoid"))

    # Compiling the CNN
    classifier.compile(
        optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"]
    )

    # Part 2 - Fitting the CNN to the images

    # Load Trained model weights

    # from keras.models import load_model
    # regressor=load_model('cnn.h5')

    train_datagen = tf.keras.preprocessing.image.ImageDataGenerator(
        rescale=1.0 / 255, shear_range=0.2, zoom_range=0.2, horizontal_flip=True
    )

    test_datagen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1.0 / 255)

    training_set = train_datagen.flow_from_directory(
        "dataset/training_set", target_size=(64, 64), batch_size=32, class_mode="binary"
    )

    test_set = test_datagen.flow_from_directory(
        "dataset/test_set", target_size=(64, 64), batch_size=32, class_mode="binary"
    )

    classifier.fit_generator(
        training_set, steps_per_epoch=5, epochs=30, validation_data=test_set
    )

    classifier.save("cnn.h5")

    # Part 3 - Making new predictions

    test_image = tf.keras.preprocessing.image.load_img(
        "dataset/single_prediction/image.png", target_size=(64, 64)
    )
    test_image = tf.keras.preprocessing.image.img_to_array(test_image)
    test_image = np.expand_dims(test_image, axis=0)
    result = classifier.predict(test_image)
    # training_set.class_indices
    if result[0][0] == 0:
        prediction = "Normal"
    if result[0][0] == 1:
        prediction = "Abnormality detected"