train_mask_detector.py

# imports
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications import MobileNetV2
from tensorflow.keras.layers import AveragePooling2D
from tensorflow.keras.layers import Dropout
from tensorflow.keras.layers import Flatten
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import Input
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
from tensorflow.keras.preprocessing.image import img_to_array
from tensorflow.keras.preprocessing.image import load_img
from tensorflow.keras.utils import to_categorical
from sklearn.preprocessing import LabelBinarizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
from imutils import paths
import matplotlib.pyplot as plt
import numpy as np
import argparse
import os
import tensorflow as tf

# Set if memory growth should be enabled for a PhysicalDevice.
physical_devices = tf.config.experimental.list_physical_devices('GPU')
for physical_device in physical_devices:
    tf.config.experimental.set_memory_growth(physical_device, True)

# argument parser construct
ap = argparse.ArgumentParser()
ap.add_argument('-d', '--dataset', required=True,
                help='path to input dataset')
ap.add_argument('-p', '--plot', type=str, default='plot.png',
                help='path to output loss/accuracy plot')
ap.add_argument('-m', '--model', type=str,
                default='mask_detector.model',
                help='path to output face mask detector model')
ap.add_argument('--cfg', type=str,
                default='config/imdim.cfg',
                help='path to image size file')
args = vars(ap.parse_args())

# Initialize the initial learning rate, number of epochs to train for,
# and batch size (Hyper params)
INIT_LR = 1e-4
EPOCHS = 20
BS = 32
IMG_RESIZE = tuple(np.flip(np.loadtxt(args['cfg']), axis=0).astype('int'))

# grab the list of images in our dataset directory, then initialize
# the list of data (i,e., images) and class images
print('[INFO] loading images...')
imagePaths = list(paths.list_images(args['dataset']))
data = []
labels = []

# loop over the image paths
for imagePath in imagePaths:
    # extract the class label from the filename
    label = imagePath.split(os.path.sep)[-2]

    # load the input image IMG_RESIZE and preprocess it
    image = load_img(imagePath, target_size=IMG_RESIZE)
    image = img_to_array(image)
    image = preprocess_input(image)

    # update the data and labels list, respectively
    data.append(image)
    labels.append(label)

# convert the data and labels to NumPy arrays
data = np.array(data, dtype='float32')
labels = np.array(labels)

# perform one-hot encoding on the labels
lb = LabelBinarizer()
labels = lb.fit_transform(labels)
labels = to_categorical(labels)

# partition the data into training and testing splits using .8 of
# the data for training an the remaining .2 for testing
(trainX, testX, trainY, testY) = train_test_split(data, labels,
                                                  test_size=.2, stratify=labels, random_state=42)

# construct the training image generator for data augmentation
aug = ImageDataGenerator(
    rotation_range=20,
    zoom_range=.15,
    width_shift_range=.2,
    height_shift_range=.2,
    shear_range=.15,
    horizontal_flip=True,
    fill_mode='nearest')

# load the MobileNetV2 network, ensuring the head FC layer sets are
# left of
baseModel = MobileNetV2(weights='imagenet', include_top=False,
                        input_tensor=Input(shape=(IMG_RESIZE[0], IMG_RESIZE[1], 3)))

# construct the head of the model that will be placed on top of
# the base model
headModel = baseModel.output
headModel = AveragePooling2D(pool_size=(7, 7))(headModel)
headModel = Flatten(name='flatten')(headModel)
headModel = Dense(128, activation='relu')(headModel)
headModel = Dropout(.5)(headModel)
headModel = Dense(2, activation='softmax')(headModel)

# place the head FC model on top of the base model (this will become
# the actual model we will train)
model = Model(inputs=baseModel.inputs, outputs=headModel)

# loop over all layers in the base model and freeze them so they will
# *not* be updated during the first traing process
for layer in baseModel.layers:
    layer.trainable = False

# compile our model
print('[INFO] compiling model...')
opt = Adam(lr=INIT_LR, decay=INIT_LR / EPOCHS)
model.compile(loss='binary_crossentropy', optimizer=opt,
              metrics=['accuracy'])

# train the head of the network
print('[INFO] training head...')
H = model.fit(
    aug.flow(trainX, trainY, batch_size=BS),
    steps_per_epoch=len(trainX) // BS,
    validation_data=(testX, testY),
    validation_steps=len(testX) // BS,
    epochs=EPOCHS)

# make predictions on the testing set
print('[INFO] evaluating network...')
predIdxs = model.predict(testX, batch_size=BS)

# for each image in the testing set we need to find the index of the
# label width corresponding largest predicted probability
predIdxs = np.argmax(predIdxs, axis=1)

# show a nicely formatted classification report
print(classification_report(testY.argmax(axis=1), predIdxs,
                            target_names=lb.classes_))

# save the model
print('[INFO] saving mask detector model...')
model.save(args['model'], save_format='h5')

# plot the training loss and accuracy
N = EPOCHS
plt.style.use('ggplot')
plt.figure()
plt.plot(np.arange(0, N), H.history['loss'], label='train_loss')
plt.plot(np.arange(0, N), H.history['val_loss'], label='val_loss')
plt.plot(np.arange(0, N), H.history['accuracy'], label='train_acc')
plt.plot(np.arange(0, N), H.history['val_accuracy'], label='val_acc')
plt.title('Training Loss and Accuracy')
plt.xlabel('Epoch #')
plt.ylabel('Loss/Accuracy')
plt.legend(loc='lower left')
plt.savefig(args['plot'])