def main(unused_argv):
logging.set_verbosity(tf.logging.INFO)
if FLAGS.use_mnist:
reader = readers.MnistReader()
else:
reader = readers.FaceReader()
if FLAGS.output_dir is "":
raise ValueError("'output_file' was not specified. "
"Unable to continue with generate_and_discriminate.")
if not os.path.exists(FLAGS.output_dir):
os.makedirs(FLAGS.output_dir)
if not os.path.exists(FLAGS.output_dir + 'images/'):
os.makedirs(FLAGS.output_dir + 'images/')
if FLAGS.input_data_pattern is "":
raise ValueError("'input_data_pattern' was not specified. "
"Unable to continue with generate_and_discriminate.")
if not (0 <= FLAGS.num_generate <= FLAGS.num_total_images):
raise ValueError("'num_generate' should be between "
"[0, num_total_images]. Unable to continue "
"with generate_and_discriminate.")
# Make separate graphs in order to load two different models.
G_graph = tf.Graph()
D_graph = tf.Graph()
images = generate(reader, G_graph, FLAGS.G_train_dir,
FLAGS.input_data_pattern, FLAGS.output_dir,
FLAGS.num_generate, FLAGS.num_total_images)
discriminate(images, D_graph, FLAGS.D_train_dir, FLAGS.output_dir)
def main(unused_argv):
logging.set_verbosity(tf.logging.INFO)
if FLAGS.use_mnist:
reader = readers.MnistReader()
else:
reader = readers.FaceReader()
if FLAGS.output_file is "":
raise ValueError("'output_file' was not specified. "
"Unable to continue with linear_interpolation.")
linear_interpolation(reader, FLAGS.train_dir, FLAGS.output_file)
def evaluate():
tf.set_random_seed(0) # for reproducibility
with tf.Graph().as_default():
if FLAGS.use_mnist:
reader = readers.MnistReader()
else:
reader = readers.FaceReader()
generator_model = find_class_by_name(FLAGS.generator_model, [models])()
discriminator_model = find_class_by_name(FLAGS.discriminator_model,
[models])()
label_loss_fn = find_class_by_name(FLAGS.label_loss, [losses])()
if FLAGS.eval_data_pattern is "":
raise IOError("'eval_data_pattern' was not specified. " +
"Nothing to evaluate.")
build_graph(reader=reader,
generator_model=generator_model,
discriminator_model=discriminator_model,
eval_data_pattern=FLAGS.eval_data_pattern,
label_loss_fn=label_loss_fn,
num_readers=FLAGS.num_readers,
batch_size=FLAGS.batch_size)
logging.info("built evaluation graph")
p_fake_batch = tf.get_collection("p_for_fake")[0]
p_real_batch = tf.get_collection("p_for_data")[0]
G_loss = tf.get_collection("G_loss")[0]
D_loss = tf.get_collection("D_loss")[0]
noise_input = tf.get_collection("noise_input_placeholder")[0]
summary_op = tf.get_collection("summary_op")[0]
saver = tf.train.Saver(tf.global_variables())
summary_writer = tf.summary.FileWriter(FLAGS.train_dir,
graph=tf.get_default_graph())
evl_metrics = eval_util.EvaluationMetrics()
last_global_step_val = -1
while True:
last_global_step_val = evaluation_loop(p_fake_batch, p_real_batch,
G_loss, D_loss, noise_input,
summary_op, saver,
summary_writer, evl_metrics,
last_global_step_val,
FLAGS.batch_size)
if FLAGS.run_once:
break
def get_reader():
if FLAGS.use_mnist:
reader = readers.MnistReader()
else:
reader = readers.FaceReader()
return reader