def __init__(self, nb_classes=2):
self.train_images = None
self.train_labels = None
self.valid_images = None
self.valid_labels = None
self.test_images = None
self.test_labels = None
self.input_shape = None
self.nb_classes = nb_classes
self.datasets = LoadData()
import random
import matplotlib.image as mpimg
import numpy as np
tf.python.control_flow_ops = tf #Fix error with TF and Keras
from load_dataset import LoadData
from model import modelClass
from preprocess_augmentation import *
if __name__ == '__main__':
#Parameters
left_stcor = 0.28
right_stcor = -0.28
imgN = 5
#Load Data img and CSV
dataO = LoadData(location='../data/driving_log.csv', validationSplit=.2)
print('data', type(dataO.data))
print('valid_data', type(dataO.valid_data))
print('train_data', type(dataO.train_data))
#Split labels
y_train, y_validation = dataO.split_labels(imgN=5,
left_stcor=0.28,
right_stcor=-0.28)
#Get model
modelO = modelClass()
#Prepare data batch
batchSize = 120
class Dataset:
def __init__(self, nb_classes=2):
self.train_images = None
self.train_labels = None
self.valid_images = None
self.valid_labels = None
self.test_images = None
self.test_labels = None
self.input_shape = None
self.nb_classes = nb_classes
self.datasets = LoadData()
def load(self, grey):
faces, genders = self.datasets.load_fbDataset(grey=grey)
# faces, genders = self.datasets.load_extra_dataset(grey=grey)
# faces, genders = self.datasets.load_extra_UTKdataset(grey=grey)
#faces, genders = self.datasets.load_extra_wikiDataset(grey=grey)
faces = np.array(faces)
genders = np.array(genders)
train_images, valid_images, train_labels, valid_labels = train_test_split(
faces, genders, test_size=0.2, random_state=0)
# train_images, valid_images, train_labels, valid_labels = train_test_split(train_images, train_labels, test_size=0.2, random_state=0)
if grey == 1:
train_images = train_images.reshape(train_images.shape[0],
self.datasets.IMAGE_SIZE,
self.datasets.IMAGE_SIZE, 1)
valid_images = valid_images.reshape(valid_images.shape[0],
self.datasets.IMAGE_SIZE,
self.datasets.IMAGE_SIZE, 1)
# test_images = test_images.reshape(test_images.shape[0], self.datasets.IMAGE_SIZE, self.datasets.IMAGE_SIZE, 3)
self.input_shape = (self.datasets.IMAGE_SIZE,
self.datasets.IMAGE_SIZE, 1)
else:
train_images = train_images.reshape(train_images.shape[0],
self.datasets.IMAGE_SIZE,
self.datasets.IMAGE_SIZE, 3)
valid_images = valid_images.reshape(valid_images.shape[0],
self.datasets.IMAGE_SIZE,
self.datasets.IMAGE_SIZE, 3)
# test_images = test_images.reshape(test_images.shape[0], self.datasets.IMAGE_SIZE, self.datasets.IMAGE_SIZE, 3)
self.input_shape = (self.datasets.IMAGE_SIZE,
self.datasets.IMAGE_SIZE, 3)
print(train_images.shape[0], "train samples")
print(valid_images.shape[0], 'valid samples')
# print(test_images.shape[0], 'test samples')
train_labels = np_utils.to_categorical(train_labels)
valid_labels = np_utils.to_categorical(valid_labels)
# test_labels = np_utils.to_categorical(test_labels)
self.nb_classes = train_labels.shape[1]
train_images = train_images.astype('float32')
valid_images = valid_images.astype('float32')
# test_images = test_images.astype('float32')
train_images /= 255
valid_images /= 255
# test_images /= 255
self.train_images = train_images
self.valid_images = valid_images
# self.test_images = test_images
self.train_labels = train_labels
self.valid_labels = valid_labels
if cuda:
generator.cuda()
discriminator.cuda()
adversarial_loss.cuda()
# Initialize weights
generator.apply(weights_init_normal)
discriminator.apply(weights_init_normal)
# Configure data loader
nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0, 8])
print('Using {} dataloader workers every process'.format(nw))
dataloader = torch.utils.data.DataLoader(
LoadData(
dped_dir=dped_dir,
dataset_size=train_size,
image_size=PATCH_SIZE,
test=False,
),
batch_size=batch_size,
shuffle=True,
num_workers=nw,
pin_memory=True,
)
# Optimizers
optimizer_G = torch.optim.Adam(generator.parameters(),
lr=learning_rate,
betas=(0.5, 0.99))
optimizer_D = torch.optim.Adam(discriminator.parameters(),
lr=learning_rate,
betas=(0.5, 0.99))