def blend_images(x, progress, resolution_schedule, num_blocks):
"""Blends images of different resolutions according to `progress`.
When training `progress` is at a stable stage for resolution r, returns
image `x` downscaled to resolution r and then upscaled to `final_resolutions`,
call it x'(r).
Otherwise when training `progress` is at a transition stage from resolution
r to 2r, returns a linear combination of x'(r) and x'(2r).
Args:
x: An image `Tensor` of NHWC format with resolution `final_resolutions`.
progress: A scalar float `Tensor` of training progress.
resolution_schedule: An object of `ResolutionSchedule`.
num_blocks: An integer of number of blocks.
Returns:
An image `Tensor` which is a blend of images of different resolutions.
"""
x_blend = []
for block_id in range(1, num_blocks + 1):
alpha = _generator_alpha(block_id, progress)
scale = resolution_schedule.scale_factor(block_id)
x_blend.append(alpha * layers.upscale(layers.downscale(x, scale), scale))
return tf.add_n(x_blend)
def blend_images(x, progress, resolution_schedule, num_blocks):
"""Blends images of different resolutions according to `progress`.
When training `progress` is at a stable stage for resolution r, returns
image `x` downscaled to resolution r and then upscaled to `final_resolutions`,
call it x'(r).
Otherwise when training `progress` is at a transition stage from resolution
r to 2r, returns a linear combination of x'(r) and x'(2r).
Args:
x: An image `Tensor` of NHWC format with resolution `final_resolutions`.
progress: A scalar float `Tensor` of training progress.
resolution_schedule: An object of `ResolutionSchedule`.
num_blocks: An integer of number of blocks.
Returns:
An image `Tensor` which is a blend of images of different resolutions.
"""
x_blend = []
for block_id in range(1, num_blocks + 1):
alpha = _generator_alpha(block_id, progress)
scale = resolution_schedule.scale_factor(block_id)
x_blend.append(alpha *
layers.upscale(layers.downscale(x, scale), scale))
return tf.add_n(x_blend)
def __call__(self, x, x_cond=None, params=None):
del params
if x_cond is not None:
x = tf.concat([x, x_cond], axis=3)
responses = []
for ii, D in enumerate(self.discriminators):
responses.append(D(x, x_cond=None)) # x_cond is already concatenated
if ii != len(self.discriminators) - 1:
x = layers.downscale(x, n=2)
return responses
def test_downscale_4d_images_returns_downscaled_images(self):
x_np = np.array(
[[[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]],
[[[0, 0, 0], [-1, -2, -3]], [[1, -2, 2], [2, 5, 3]]]],
dtype=np.float32)
with self.test_session(use_gpu=True) as sess:
x1_np, x2_np = sess.run(
[layers.downscale(tf.constant(x_np), n) for n in [1, 2]])
expected2_np = [[[[5.5, 6.5, 7.5]]], [[[0.5, 0.25, 0.5]]]]
self.assertNDArrayNear(x1_np, x_np, 1.0e-5)
self.assertNDArrayNear(x2_np, expected2_np, 1.0e-5)
def test_blend_images_in_transition_stage(self):
x_np = np.random.normal(size=[2, 8, 8, 3])
x = tf.constant(x_np, tf.float32)
x_blend = networks.blend_images(
x,
tf.constant(0.2),
resolution_schedule=networks.ResolutionSchedule(scale_base=2,
num_resolutions=2),
num_blocks=2)
with self.test_session(use_gpu=True) as sess:
x_blend_np = sess.run(x_blend)
x_blend_expected_np = 0.8 * sess.run(
layers.upscale(layers.downscale(x, 2), 2)) + 0.2 * x_np
self.assertNDArrayNear(x_blend_np, x_blend_expected_np, 1.0e-6)
def discriminator(x,
progress,
num_filters_fn,
resolution_schedule,
num_blocks=None,
kernel_size=3,
scope='progressive_gan_discriminator',
reuse=None):
"""Discriminator network for the progressive GAN model.
Args:
x: A `Tensor`of NHWC format representing images of size `resolution`.
progress: A scalar float `Tensor` of training progress.
num_filters_fn: A function that maps `block_id` to # of filters for the
block.
resolution_schedule: An object of `ResolutionSchedule`.
num_blocks: An integer of number of blocks. None means maximum number of
blocks, i.e. `resolution.schedule.num_resolutions`. Defaults to None.
kernel_size: An integer of convolution kernel size.
scope: A string or variable scope.
reuse: Whether to reuse `scope`. Defaults to None which means to inherit
the reuse option of the parent scope.
Returns:
A `Tensor` of model output and a dictionary of model end points.
"""
if num_blocks is None:
num_blocks = resolution_schedule.num_resolutions
def _conv2d(scope, x, kernel_size, filters, padding='SAME'):
return layers.custom_conv2d(x=x,
filters=filters,
kernel_size=kernel_size,
padding=padding,
activation=tf.nn.leaky_relu,
he_initializer_slope=0.0,
scope=scope)
def _from_rgb(x, block_id):
return _conv2d('from_rgb', x, 1, num_filters_fn(block_id))
end_points = {}
with tf.variable_scope(scope, reuse=reuse):
x0 = x
end_points['rgb'] = x0
lods = []
for block_id in range(num_blocks, 0, -1):
with tf.variable_scope(block_name(block_id)):
scale = resolution_schedule.scale_factor(block_id)
lod = layers.downscale(x0, scale)
end_points['downscaled_rgb_{}'.format(block_id)] = lod
lod = _from_rgb(lod, block_id)
# alpha_i is used to replace lod_select.
alpha = _discriminator_alpha(block_id, progress)
end_points['alpha_{}'.format(block_id)] = alpha
lods.append((lod, alpha))
lods_iter = iter(lods)
x, _ = lods_iter.__next__()
for block_id in range(num_blocks, 1, -1):
with tf.variable_scope(block_name(block_id)):
x = _conv2d('conv0', x, kernel_size, num_filters_fn(block_id))
x = _conv2d('conv1', x, kernel_size,
num_filters_fn(block_id - 1))
x = layers.downscale(x, resolution_schedule.scale_base)
lod, alpha = lods_iter.__next__()
x = alpha * lod + (1.0 - alpha) * x
with tf.variable_scope(block_name(1)):
x = layers.scalar_concat(x, layers.minibatch_mean_stddev(x))
x = _conv2d('conv0', x, kernel_size, num_filters_fn(1))
x = _conv2d('conv1', x, resolution_schedule.start_resolutions,
num_filters_fn(0), 'VALID')
end_points['last_conv'] = x
logits = layers.custom_dense(x=x, units=1, scope='logits')
end_points['logits'] = logits
return logits, end_points
def test_downscale_invalid_scale_throws_exception(self):
with self.assertRaises(ValueError):
layers.downscale(tf.constant([]), -1)
def discriminator(x,
progress,
num_filters_fn,
resolution_schedule,
num_blocks=None,
kernel_size=3,
scope='progressive_gan_discriminator',
reuse=None):
"""Discriminator network for the progressive GAN model.
Args:
x: A `Tensor`of NHWC format representing images of size `resolution`.
progress: A scalar float `Tensor` of training progress.
num_filters_fn: A function that maps `block_id` to # of filters for the
block.
resolution_schedule: An object of `ResolutionSchedule`.
num_blocks: An integer of number of blocks. None means maximum number of
blocks, i.e. `resolution.schedule.num_resolutions`. Defaults to None.
kernel_size: An integer of convolution kernel size.
scope: A string or variable scope.
reuse: Whether to reuse `scope`. Defaults to None which means to inherit
the reuse option of the parent scope.
Returns:
A `Tensor` of model output and a dictionary of model end points.
"""
if num_blocks is None:
num_blocks = resolution_schedule.num_resolutions
def _conv2d(scope, x, kernel_size, filters, padding='SAME'):
return layers.custom_conv2d(
x=x,
filters=filters,
kernel_size=kernel_size,
padding=padding,
activation=tf.nn.leaky_relu,
he_initializer_slope=0.0,
scope=scope)
def _from_rgb(x, block_id):
return _conv2d('from_rgb', x, 1, num_filters_fn(block_id))
end_points = {}
with tf.variable_scope(scope, reuse=reuse):
x0 = x
end_points['rgb'] = x0
lods = []
for block_id in range(num_blocks, 0, -1):
with tf.variable_scope(block_name(block_id)):
scale = resolution_schedule.scale_factor(block_id)
lod = layers.downscale(x0, scale)
end_points['downscaled_rgb_{}'.format(block_id)] = lod
lod = _from_rgb(lod, block_id)
# alpha_i is used to replace lod_select.
alpha = _discriminator_alpha(block_id, progress)
end_points['alpha_{}'.format(block_id)] = alpha
lods.append((lod, alpha))
lods_iter = iter(lods)
x, _ = lods_iter.next()
for block_id in range(num_blocks, 1, -1):
with tf.variable_scope(block_name(block_id)):
x = _conv2d('conv0', x, kernel_size, num_filters_fn(block_id))
x = _conv2d('conv1', x, kernel_size, num_filters_fn(block_id - 1))
x = layers.downscale(x, resolution_schedule.scale_base)
lod, alpha = lods_iter.next()
x = alpha * lod + (1.0 - alpha) * x
with tf.variable_scope(block_name(1)):
x = layers.scalar_concat(x, layers.minibatch_mean_stddev(x))
x = _conv2d('conv0', x, kernel_size, num_filters_fn(1))
x = _conv2d('conv1', x, resolution_schedule.start_resolutions,
num_filters_fn(0), 'VALID')
end_points['last_conv'] = x
logits = layers.custom_dense(x=x, units=1, scope='logits')
end_points['logits'] = logits
return logits, end_points