def test_single_example_correct(self, mock_tfhub_load):
mock_tfhub_load.return_value = fake_logit_fn
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 2, 0),
tf.concat([real_digit()] * 2, 0))
with self.cached_session() as sess:
self.assertNear(0.0, sess.run(fdistance), 2e-1)
def calculate_fid(opts):
model_path = opts['--model-path']
model = get_generator_model()
model = load_model(model_path)
(_, _), (x_test, _) = mnist.load_data()
x_test = (x_test.astype(numpy.float32)) / 255.0
x_test = numpy.expand_dims(x_test, axis=3)
random_code = numpy.random.normal(size=(NUM_PICS_TO_CALCULATE_FID,
LATENT_DIM))
predicted_imgs = model.predict(random_code)
numpy.random.shuffle(x_test)
real_imgs = x_test[:NUM_PICS_TO_CALCULATE_FID]
predicted_imgs = tf.convert_to_tensor(predicted_imgs, dtype=tf.float32)
real_imgs = tf.convert_to_tensor(real_imgs, dtype=tf.float32)
fid = util.mnist_frechet_distance(real_imgs, predicted_imgs)
print("CALCULATED FID: %f" % fid)
if opts.get('--output', None):
path = opts['--output']
if os.path.exists(path):
os.remove(path)
with open(path, "w") as f:
f.write("CALCULATED FID: %f" % fid)
def test_batch_splitting_doesnt_change_value(self):
for num_batches in [1, 2, 4, 8]:
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 6 + [fake_digit()] * 2, 0),
tf.concat([real_digit()] * 2 + [fake_digit()] * 6, 0),
num_batches=num_batches)
with self.cached_session() as sess:
self.assertNear(97.8, sess.run(fdistance), 2e-1)
def test_minibatch_correct(self):
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit(), real_digit(),
fake_digit()], 0),
tf.concat([real_digit(), fake_digit(),
fake_digit()], 0))
with self.cached_session() as sess:
self.assertNear(43.5, sess.run(fdistance), 2e-1)
def _disabled_test_minibatch_correct(self):
"""Tests the correctness of the mnist_frechet_distance function."""
# Disabled since it requires loading the tfhub MNIST module.
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit(), real_digit(), fake_digit()], 0),
tf.concat([real_digit(), fake_digit(), fake_digit()], 0))
with self.cached_session() as sess:
self.assertNear(43.5, sess.run(fdistance), 2e-1)
def test_single_example_correct(self):
if tf.executing_eagerly():
# `run_image_classifier` doesn't work in eager.
return
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 2, 0),
tf.concat([real_digit()] * 2, 0))
with self.cached_session() as sess:
self.assertNear(0.0, sess.run(fdistance), 2e-1)
def test_minibatch_correct(self):
if tf.executing_eagerly():
# `run_image_classifier` doesn't work in eager.
return
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit(), real_digit(), fake_digit()], 0),
tf.concat([real_digit(), fake_digit(), fake_digit()], 0))
with self.cached_session() as sess:
self.assertNear(43.5, sess.run(fdistance), 2e-1)
def test_any_batch_size(self):
if tf.executing_eagerly():
# Placeholders don't work in eager execution mode.
return
inputs = tf.compat.v1.placeholder(tf.float32, shape=[None, 28, 28, 1])
fdistance = util.mnist_frechet_distance(inputs, inputs)
for batch_size in [4, 16, 30]:
with self.cached_session() as sess:
sess.run(fdistance,
feed_dict={inputs: np.zeros([batch_size, 28, 28, 1])})
def test_any_batch_size(self, mock_tfhub_load):
mock_tfhub_load.return_value = fake_logit_fn
# Create a graph since placeholders don't work in eager execution mode.
with tf.Graph().as_default():
inputs = tf.placeholder(tf.float32, shape=[None, 28, 28, 1])
fdistance = util.mnist_frechet_distance(inputs, inputs)
for batch_size in [4, 16, 30]:
with self.cached_session() as sess:
sess.run(fdistance,
feed_dict={inputs: np.zeros([batch_size, 28, 28, 1])})
def get_metrics(gan_model):
"""Return metrics for MNIST experiment."""
real_mnist_score = util.mnist_score(gan_model.real_data)
generated_mnist_score = util.mnist_score(gan_model.generated_data)
frechet_distance = util.mnist_frechet_distance(gan_model.real_data,
gan_model.generated_data)
return {
'real_mnist_score': tf.compat.v1.metrics.mean(real_mnist_score),
'mnist_score': tf.compat.v1.metrics.mean(generated_mnist_score),
'frechet_distance': tf.compat.v1.metrics.mean(frechet_distance),
}
def test_deterministic(self):
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 2, 0), tf.concat([fake_digit()] * 2, 0))
with self.cached_session() as sess:
fdistance1 = sess.run(fdistance)
fdistance2 = sess.run(fdistance)
self.assertNear(fdistance1, fdistance2, 2e-1)
with self.cached_session() as sess:
fdistance3 = sess.run(fdistance)
self.assertNear(fdistance1, fdistance3, 2e-1)
def _disabled_test_batch_splitting_doesnt_change_value(self):
"""Tests correctness of mnist_frechet_distance function with batch sizes."""
# Disabled since it requires loading the tfhub MNIST module.
with tf.Graph().as_default():
for num_batches in [1, 2, 4, 8]:
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 6 + [fake_digit()] * 2, 0),
tf.concat([real_digit()] * 2 + [fake_digit()] * 6, 0),
num_batches=num_batches)
with self.cached_session() as sess:
self.assertNear(97.8, sess.run(fdistance), 2e-1)
def test_batch_splitting_doesnt_change_value(self):
if tf.executing_eagerly():
# `run_image_classifier` doesn't work in eager.
return
for num_batches in [1, 2, 4, 8]:
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 6 + [fake_digit()] * 2, 0),
tf.concat([real_digit()] * 2 + [fake_digit()] * 6, 0),
num_batches=num_batches)
with self.cached_session() as sess:
self.assertNear(97.8, sess.run(fdistance), 2e-1)
def test_deterministic(self, mock_tfhub_load):
mock_tfhub_load.return_value = fake_logit_fn
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 2, 0), tf.concat([fake_digit()] * 2, 0))
with self.cached_session() as sess:
fdistance1 = sess.run(fdistance)
fdistance2 = sess.run(fdistance)
self.assertNear(fdistance1, fdistance2, 2e-1)
with self.cached_session() as sess:
fdistance3 = sess.run(fdistance)
self.assertNear(fdistance1, fdistance3, 2e-1)
def evaluate(hparams, run_eval_loop=True):
"""Runs an evaluation loop.
Args:
hparams: An HParams instance containing the eval hyperparameters.
run_eval_loop: Whether to run the full eval loop. Set to False for testing.
"""
# Fetch real images.
with tf.compat.v1.name_scope('inputs'):
real_images, _ = data_provider.provide_data(
'train', hparams.num_images_generated, hparams.dataset_dir)
image_write_ops = None
if hparams.eval_real_images:
tf.compat.v1.summary.scalar(
'MNIST_Classifier_score',
util.mnist_score(real_images, hparams.classifier_filename))
else:
# In order for variables to load, use the same variable scope as in the
# train job.
with tf.compat.v1.variable_scope('Generator'):
images = networks.unconditional_generator(tf.random.normal(
[hparams.num_images_generated, hparams.noise_dims]),
is_training=False)
tf.compat.v1.summary.scalar(
'MNIST_Frechet_distance',
util.mnist_frechet_distance(real_images, images,
hparams.classifier_filename))
tf.compat.v1.summary.scalar(
'MNIST_Classifier_score',
util.mnist_score(images, hparams.classifier_filename))
if hparams.num_images_generated >= 100 and hparams.write_to_disk:
reshaped_images = tfgan.eval.image_reshaper(images[:100, ...],
num_cols=10)
uint8_images = data_provider.float_image_to_uint8(reshaped_images)
image_write_ops = tf.io.write_file(
'%s/%s' % (hparams.eval_dir, 'unconditional_gan.png'),
tf.image.encode_png(uint8_images[0]))
# For unit testing, use `run_eval_loop=False`.
if not run_eval_loop:
return
evaluation.evaluate_repeatedly(
hparams.checkpoint_dir,
hooks=[
evaluation.SummaryAtEndHook(hparams.eval_dir),
evaluation.StopAfterNEvalsHook(1)
],
eval_ops=image_write_ops,
max_number_of_evaluations=hparams.max_number_of_evaluations)
def test_deterministic(self):
if tf.executing_eagerly():
# `run_image_classifier` doesn't work in eager.
return
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 2, 0),
tf.concat([fake_digit()] * 2, 0))
with self.cached_session() as sess:
fdistance1 = sess.run(fdistance)
fdistance2 = sess.run(fdistance)
self.assertNear(fdistance1, fdistance2, 2e-1)
with self.cached_session() as sess:
fdistance3 = sess.run(fdistance)
self.assertNear(fdistance1, fdistance3, 2e-1)
def get_mnist_eval_metrics(real, fake):
frechet = tfgan_mnist.mnist_frechet_distance(real, fake, 1)
score = tfgan_mnist.mnist_score(fake, 1)
return tf.stack(list(map(tf.stop_gradient, (frechet, score))))
def test_single_example_correct(self):
fdistance = util.mnist_frechet_distance(
tf.concat([real_digit()] * 2, 0), tf.concat([real_digit()] * 2, 0))
with self.cached_session() as sess:
self.assertNear(0.0, sess.run(fdistance), 2e-1)
