def main():
# Parse the CLI arguments.
args = parser.parse_args()
# create directory for saving trained models.
if not os.path.exists('models'):
os.makedirs('models')
# Create the tensorflow dataset.
ds = DataLoader(args.input_dir, args.target_dir,
args.image_size).dataset(args.batch_size)
# Initialize the GAN object.
gan = FastSRGAN(args)
if args.dis != None and args.gen != None:
print("Loading pre-trained generator...")
gan.generator = keras.models.load_model(args.gen)
print("Loading pre-trained discriminator...")
gan.discriminator = keras.models.load_model(args.dis)
else:
# Define the directory for saving pretrainig loss tensorboard summary.
pretrain_summary_writer = tf.summary.create_file_writer(
'logs/pretrain')
# Run pre-training.
pretrain_generator(gan, ds, pretrain_summary_writer)
# Define the directory for saving the SRGAN training tensorbaord summary.
train_summary_writer = tf.summary.create_file_writer('logs/train')
# Run training.
for epoch in range(args.epochs):
print("Training Epoch #", epoch)
train(gan, ds, args.save_iter, train_summary_writer)
def main():
# 分析/处理CLI参数.
args = parser.parse_args()
# 创建目录以存储训练的模型
if not os.path.exists('models'):
os.makedirs('models')
# 创建tensorflow数据集
ds = DataLoader(args.image_dir, args.hr_size).dataset(args.batch_size)
# 初始化GAN对象
gan = FastSRGAN(args)
# 定义保存预训练loss tensorboard摘要的目录
pretrain_summary_writer = tf.summary.create_file_writer('logs/pretrain')
# 执行预训练
pretrain_generator(gan, ds, pretrain_summary_writer)
# 定义保存SRGAN训练tensorbaord摘要的目录
train_summary_writer = tf.summary.create_file_writer('logs/train')
# 执行训练
for _ in range(args.epochs):
train(gan, ds, args.save_iter, train_summary_writer)
def main():
# Parse the CLI arguments.
args = parser.parse_args()
# create directory for saving trained models.
if not os.path.exists('models'):
os.makedirs('models')
# Create the tensorflow dataset.
ds = DataLoader(args.image_dir, args.hr_size).dataset(args.batch_size)
# Initialize the GAN object.
gan = FastSRGAN(args)
# Define the directory for saving pretrainig loss tensorboard summary.
pretrain_summary_writer = tf.summary.create_file_writer('logs/pretrain')
# Run pre-training.
pretrain_generator(gan, ds, pretrain_summary_writer)
# Define the directory for saving the SRGAN training tensorbaord summary.
train_summary_writer = tf.summary.create_file_writer('logs/train')
# Run training.
for _ in range(args.epochs):
train(gan, ds, args.save_iter, train_summary_writer)
def main():
# Parse the CLI arguments.
args = parser.parse_args()
# create directory for saving trained models.
if not os.path.exists('models'):
os.makedirs('models')
# Create the tensorflow dataset.
print('Dataset Loading.....')
ds = DataLoader(args.image_dir, args.hr_size).dataset(args.batch_size)
print('Complete!')
# Initialize the GAN object.
print('GAN Initializing......')
gan = FastSRGAN(args)
print('Complete!')
# Define the directory for saving pretrainig loss tensorboard summary.
print('Pretraon_summary_writing.....')
pretrain_summary_writer = tf.summary.create_file_writer('logs/pretrain')
print('Complete!')
# Run pre-training.
print('Run Pre-training.....')
pretrain_generator(gan, ds, pretrain_summary_writer)
print('Complete!')
# Define the directory for saving the SRGAN training tensorbaord summary.
train_summary_writer = tf.summary.create_file_writer('logs/train')
# Run training.
print('===== Training Start =====')
for i in range(args.epochs):
print('epoch = {:02d}/{:02d}'.format(i+1, args.epochs))
train(gan, ds, args.save_iter, train_summary_writer)
print('Done!')