model_before_modularkeras.py

import csv
import cv2
import numpy as np
import matplotlib.pyplot as plt
import sklearn
from sklearn.model_selection import train_test_split
import utils
from PIL import Image
import math
import keras_model
from keras_model import KerasModel
from random import shuffle


def augment_data(images, measurements):
    augmented_images = []  # + images
    augmented_measurements = []  # + measurements
    for image, steering_angle in zip(images, measurements):
        flipped_image, flipped_steering_angle = flip_image_steering(image, steering_angle)
        augmented_images.append(flipped_image)
        augmented_images.append(image)
        augmented_measurements.append(flipped_steering_angle)
        augmented_measurements.append(steering_angle)
    # utils.beep()
    return augmented_images, augmented_measurements


def flip_image_steering(image, steering_angle):
    flipped_image = np.fliplr(image)
    flipped_steering_angle = steering_angle * -1.0
    return flipped_image, flipped_steering_angle


def load_csv_data(file_path):
    # Reading the recorded data from the .csv file
    lines = []
    with open(file_path) as csvfile:
        reader = csv.reader(csvfile)
        for line in reader:
            lines.append(line)
    return lines


def get_images_and_measurements(lines):
    images = []
    measurements = []

    for line in lines:
        steering_center_angel = float(line[3])

        # Adjusted Steering angels for side camera images
        correction_factor = 0.2  # should change this to a computed parameter
        steering_left_angel = steering_center_angel + correction_factor
        steering_right_angel = steering_center_angel - correction_factor

        # Read in the images form center, left, and right cameras
        center_image = np.asarray(Image.open(line[0]))
        left_image = np.asarray(Image.open(line[1]))
        right_image = np.asarray(Image.open(line[2]))
        # np.asarray(Image.open(path + row[0]))
        # image = cv2.imread(source_path)

        # Add images and angels to the dataset
        images.extend([center_image, left_image, right_image])
        measurements.extend([steering_center_angel, steering_left_angel, steering_right_angel])
    return images, measurements

#  Keras Data generator


# class DataGenerator(keras.utils.Sequence):
#     'This class generates data for keras fit_generator()'
#     N_CAMERA_IMAGES = 3
#     N_AUGMENTATION = 1 + 1
#
#     def __int__(self,
#                 samples,
#                 batch_size,
#                 dim, n_channels,
#                 n_classes,
#                 shuffle=True,
#                 validation=False):
#         self.dim = dim
#         self.batch_size = batch_size
#         self.samples = samples
#         self.n_channels = n_channels
#         self.n_classes = n_classes
#         self.shuffle = shuffle
#         self.on_epoch_end()
#
#     def on_epoch_end(self):
#         'Shuffle the data after each epoch'
#         if self.shuffle:
#             shuffle(self.samples)
#
#     def __data_generation(self, index):
#         'Generate batch data'
#
#     def __len__(self):
#         'Gives the number of batches that Keras fit_generator expects before moving on to the next epoch'
#         return math.ceil((len(self.samples) * N_CAMERA_IMAGES * N_AUGMENTATION) / self.batch_size)
#
#     def __getitem__(self, index):
#         'Generates a batch'
#         if index + 1 == len(self):
#             X_data, y_data = self.__data_generation
#         else:
#             start_index = math.ceil((index * self.batch_size) / (N_CAMERA_IMAGES * N_AUGMENTATION))
#             end_index = math.ceil(((index + 1) * self.batch_size) /
#                                   (N_CAMERA_IMAGES * N_AUGMENTATION))
#             n_samples_to_generate = math.ceil(
#                 (self.batch_size * (index + 1)) / N_CAMERA_IMAGES * N_AUGMENTATION)
#
#
# def data_generator(samples, batch_size):
#     print('here')
#     n_samples = len(samples) * 3 * 2  # number of samples * 3 camera images * augmentation
#     # print('Number of samples')
#     while 1:  # Forever loop to keep the generator up till the termination of the program
#              # (end of training and inference)
#         # Shuffle the data before bedfore batching batch data
#         # if n_samples == 29880:
#         #     print('\nTraining\n===========================')
#         # else:
#         #     print('\nValidation\n=========================')
#         # print('Number of samples {}'.format(n_samples))
#         shuffle(samples)
#         for offset in range(0, n_samples, batch_size):
#             # Create batch of batch_size
#             batch_samples = samples[offset: offset + batch_size]
#             # Get images and measurements (angels) for the batch
#             batch_images, batch_measurements = get_images_and_measurements(batch_samples)
#             # Augment the batch dataset
#             augmented_batch_images, augmented_batch_measurements = augment_data(batch_images,
#                                                                                 batch_measurements)
#             # Putting our augmented data into numpy arrays cause Keras require numpy arrays
#             batch_features = np.array(augmented_batch_images)
#             batch_labels = np.array(augmented_batch_measurements)
#             # Shuffle the batch data for good measure
#             print(' X_train: {} and y_train: {}'.format(batch_features.shape, batch_labels.shape))
#             yield shuffle(batch_features, batch_labels)


def data_generator1(samples, batch_size, get_number=False):
    print('here')
    while True:  # Forever loop to keep the generator up till the termination of the program
             # (end of training and inference)
        shuffle(samples)
        X_data = []
        y_data = []
        for i, sample in enumerate(samples):
            # Get the samples images, which will return 3 images (center, left, right)
            # and their angles
            sample_images, sample_measurements = get_images_and_measurements([sample])
            # Augment sample images (flip)
            augmented_sample_images, augmented_sample_measurements = augment_data(sample_images,
                                                                                  sample_measurements)
            # Adding our generated sample data into our yield arrays
            X_data.extend(augmented_sample_images)
            y_data.extend(augmented_sample_measurements)
            # print('X_data length: {}'.format(len(X_data)))
            # Check if X is of batch_size or if its the last element
            if len(X_data) > batch_size or i == len(samples) - 1:
                # print('==================Batch====================')
                # Putting our augmented data into numpy arrays cause Keras require numpy arrays
                # yield the batch
                # Shuffle the batch data for good measure
                yield sklearn.utils.shuffle(np.array(X_data[:batch_size]), np.array(y_data[:batch_size]))
                X_data = X_data[batch_size:]
                y_data = y_data[batch_size:]


def plot_loss(model_history):
    print(model_history.history.keys())
    plt.plot(model_history.history['loss'])
    plt.plot(model_history.history['val_loss'])
    plt.title('model mean squared error loss')
    plt.ylabel('mean squared error loss')
    plt.xlabel('epoch')
    plt.legend(['training set', 'validation set'], loc='upper right')
    plt.show()


# Hyperparameters
EPOCHS = 20
BATCHSIZE = 512

# Load data from csv file
lines = load_csv_data('./DrivingData/driving_log.csv')
# lines = load_csv_data('./DrivingData_track2/driving_log.csv')
# Splitting the data to a 80% training and 20% validation
train_samples, validation_samples = train_test_split(lines, test_size=0.2)
# print('train_samples {}, validation_samples {}'.format(len(train_samples), len(validation_samples)))
train_generator = data_generator1(train_samples, batch_size=BATCHSIZE)
validation_generator = data_generator1(validation_samples, batch_size=BATCHSIZE)

# # Keras LeNet Model
# model = Sequential()
# # Normalizing and standardizing our images
# model.add(Lambda(lambda x: x / 255.0 - 0.5, input_shape=(160, 320, 3)))
# # Cropping our images using Cropping2D
# model.add(Cropping2D(cropping=((70, 25), (0, 0))))
# # First Convolution2D layer with
# model.add(Convolution2D(6, (5, 5), activation='relu'))
# # MaxPooling2D layer
# model.add(MaxPooling2D())
# # Second Convolution2D layer with
# model.add(Convolution2D(6, (5, 5), activation='relu'))
# # MaxPooling2D layer
# model.add(MaxPooling2D())
# # Flattening the Images after the convolutional steps
# model.add(Flatten())
# # Fist dense layer
# model.add(Dense(120))
# # Second dense layer
# model.add(Dense(84))
# # Logits layer
# model.add(Dense(1))
# # Defining the loss function and optimizer
# model.compile(loss='mse', optimizer='adam')

training_lenght = math.ceil((len(train_samples)*3*2) / BATCHSIZE)
validation_length = math.ceil((len(validation_samples)*3*2) / BATCHSIZE)
# print(len(list(train_generator)))
k_model = KerasModel(1, keras_model.LENET_ARCHITECTURE)
model_history = k_model.train_model_with_generator(train_generator,
                                                   training_lenght,
                                                   EPOCHS,
                                                   validation_generator,
                                                   validation_length,
                                                   save_model_filepath='model_modular.h5')
# model_history = model.fit_generator(train_generator,
#                                     steps_per_epoch=training_lenght,
#                                     validation_data=validation_generator,
#                                     validation_steps=validation_length,
#                                     epochs=EPOCHS, verbose=1)
#
# model.save('model.h5')
# model.save('model_track2.h5')
plot_loss(model_history=model_history)