summary_writer.add_image("UV", dataset.uv, channels=2)
summary_op = tf.summary.merge(summary_writer.lists)
return Model(train_op=train_op,
summary_op=summary_op,
loss=loss,
vars=tf_vars,
output=reconstruct_mask)
if __name__ == '__main__':
set_random_seed(args)
# setting up logging
logger = initial_logger(args)
args.logger = logger
args.ngf = 32
args.resnet_conv_count = 2
args.resnet_res_count = 9
args.resnet_padding = 'SYMMETRIC'
# load data and preprocess
dataset = data_mask.load_data(args)
# build network
logger.info('------Build Network Structure------')
model = create_model(dataset, args)
exit(-1)
def create_test_model(uv, args):
with tf.variable_scope("mask_branch"):
reconstruct_mask = resnet_cyclegan(uv,
output_channles=1,
activation="sigmoid",
prefix="mask",
args=args)
return reconstruct_mask
if __name__ == '__main__':
# setting up logging
logger = initial_logger(args, dump_code=False)
args.logger = logger
args.resolution = 512
args.ngf = 32
args.resnet_conv_count = 2
args.resnet_res_count = 9
args.resnet_padding = 'SYMMETRIC'
args.batch_size = 1
# build network
logger.info('------Build Network Structure------')
tf_uv = tf.placeholder(dtype=tf.float32,
shape=[1, args.resolution, args.resolution, 2])
output = create_test_model(tf_uv, args)