居眠りをするかどうか

参照先:https://github.com/AnirudhGP/DrowsyDriverDetection

目を閉じて

目を前処理し、対応する検証セットとテストセットを得てpickleで保存します.

import matplotlib.pyplot as plt
import numpy as np
import os

from six.moves import cPickle as pickle
import cv2

dirs = ['Dataset/openLeftEyes', 'Dataset/openRightEyes']
dirs2 = ['Dataset/closedLeftEyes', 'Dataset/closedRightEyes']


def generate_dataset():
    dataset = np.ndarray([1231 * 2, 24, 24, 1], dtype='float32')
    i = 0
    for dir in dirs:
        for filename in os.listdir(dir):
            if filename.endswith('.jpg'):
                im = cv2.imread(dir + '/' + filename)
                im = np.dot(np.array(im, dtype='float32'), [[0.2989], [0.5870], [0.1140]]) / 255
                dataset[i, :, :, :] = im[:, :, :]
                i += 1
    labels = np.ones([len(dataset), 1], dtype=int)
    return dataset, labels


def generate_dataset_closed():
    dataset = np.ndarray([1192 * 2, 24, 24, 1], dtype='float32')
    i = 0
    for dir in dirs2:
        for filename in os.listdir(dir):
            if filename.endswith('.jpg'):
                im = cv2.imread(dir + '/' + filename)
                im = np.dot(np.array(im, dtype='float32'), [[0.2989], [0.5870], [0.1140]]) / 255
                dataset[i, :, :, :] = im[:, :, :]
                i += 1
    labels = np.zeros([len(dataset), 1], dtype=int)
    return dataset, labels


dataset_open, labels_open = generate_dataset()
dataset_closed, labels_closed = generate_dataset_closed()
print("done")

split = int(len(dataset_closed) * 0.8)
train_dataset_closed = dataset_closed[:split]
train_labels_closed = labels_closed[:split]
test_dataset_closed = dataset_closed[split:]
test_labels_closed = labels_closed[split:]

pickle_file = 'closed_eyes.pickle'

try:
    f = open(pickle_file, 'wb')
    save = {
     
        'train_dataset': train_dataset_closed,
        'train_labels': train_labels_closed,
        'test_dataset': test_dataset_closed,
        'test_labels': test_labels_closed,
    }
    pickle.dump(save, f, pickle.HIGHEST_PROTOCOL)
    f.close()
except Exception as e:
    print('Unable to save data to', pickle_file, ':', e)
    raise

statinfo = os.stat(pickle_file)
print('Compressed pickle size:', statinfo.st_size)

split = int(len(dataset_open) * 0.8)
train_dataset_open = dataset_open[:split]
train_labels_open = labels_open[:split]
test_dataset_open = dataset_open[split:]
test_labels_open = labels_open[split:]

pickle_file = 'open_eyes.pickle'

try:
    f = open(pickle_file, 'wb')
    save = {
     
        'train_dataset': train_dataset_open,
        'train_labels': train_labels_open,
        'test_dataset': test_dataset_open,
        'test_labels': test_labels_open,
    }
    pickle.dump(save, f, pickle.HIGHEST_PROTOCOL)
    f.close()
except Exception as e:
    print('Unable to save data to', pickle_file, ':', e)
    raise

statinfo = os.stat(pickle_file)
print('Compressed pickle size:', statinfo.st_size)

コンボリューションニューラルネットワークで訓練する

from __future__ import absolute_import
from __future__ import print_function
import numpy as np
import os
from keras.utils import plot_model
import cv2

np.random.seed(1337)  # for reproducibility



from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation, Flatten
from keras.layers.convolutional import Convolution2D, MaxPooling2D
from keras.utils import np_utils
from keras.optimizers import SGD, Adadelta, Adagrad

from six.moves import cPickle as pickle

import tensorflow as tf
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))

# os.environ["PATH"] += os.pathsep + 'C:/Program Files (x86)/Graphviz2.38/bin/'

pickle_files = ['open_eyes.pickle', 'closed_eyes.pickle']
i = 0
for pickle_file in pickle_files:
    with open(pickle_file, 'rb') as f:
        save = pickle.load(f)
        if i == 0:
            train_dataset = save['train_dataset']
            train_labels = save['train_labels']
            test_dataset = save['test_dataset']
            test_labels = save['test_labels']
        else:
            print("here")
            train_dataset = np.concatenate((train_dataset, save['train_dataset']))
            train_labels = np.concatenate((train_labels, save['train_labels']))
            test_dataset = np.concatenate((test_dataset, save['test_dataset']))
            test_labels = np.concatenate((test_labels, save['test_labels']))
        del save  # hint to help gc free up memory
    i += 1

print('Training set', train_dataset.shape, train_labels.shape)
print('Test set', test_dataset.shape, test_labels.shape)

batch_size = 30
nb_classes = 1
epochs = 12

X_train = train_dataset
X_train = X_train.reshape((X_train.shape[0], X_train.shape[3]) + X_train.shape[1:3]) #(3876, 1, 24, 24)
Y_train = train_labels

X_test = test_dataset
X_test = X_test.reshape((X_test.shape[0], X_test.shape[3]) + X_test.shape[1:3])
Y_test = test_labels

# print shape of data while model is building
print("{1} train samples, {4} channel{0}, {2}x{3}".format("" if X_train.shape[1] == 1 else "s", *X_train.shape))
print("{1}  test samples, {4} channel{0}, {2}x{3}".format("" if X_test.shape[1] == 1 else "s", *X_test.shape))

# input image dimensions
_, img_channels, img_rows, img_cols = X_train.shape

# convert class vectors to binary class matrices
# Y_train = np_utils.to_categorical(y_train, nb_classes)
# Y_test = np_utils.to_categorical(y_test, nb_classes)

model = Sequential()

model.add(Convolution2D(32, (3, 3), padding='same',
                        input_shape=(img_channels, img_rows, img_cols),data_format='channels_first'))
model.add(Activation('relu'))
model.add(Convolution2D(24, (3, 3), data_format='channels_first'),)
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Convolution2D(64, (3, 3), padding='same', data_format='channels_first'))
model.add(Activation('relu'))
model.add(Convolution2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes))
model.add(Activation('sigmoid'))

# let's train the model using SGD + momentum (how original).
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='binary_crossentropy', optimizer=sgd, metrics=['accuracy'])

model.fit(X_train, Y_train, batch_size=batch_size, epochs=epochs, verbose=2, validation_data=(X_test, Y_test))


model.save('eye_model.h5')

score = model.evaluate(X_test, Y_test,  verbose=1)

print('Test score:', score[0])
print('Test accuracy:', score[1])

テストコード

#test code
import cv2
import numpy as np
im = cv2.imread('test.jpg')
im = np.dot(np.array(im, dtype='float32'), [[0.2989], [0.5870], [0.1140]]) / 255
im=im.reshape(1,24,24)
im=np.expand_dims(im,0)
print(im.shape)
#
#model.predict(im, batch_size=1, verbose=1, steps=None)

from keras.models import load_model
my_model = load_model('eye_model.h5')
pre=my_model.predict(im)
print(pre)