def create_legacy_encoder(inputs,
nz=8,
nef=64,
norm_layer='instance',
include_top=True):
norm_layer = ops.get_norm_layer(norm_layer)
with tf.variable_scope('h1'):
h1 = conv_pool2d(inputs, nef, kernel_size=5, strides=2)
h1 = norm_layer(h1)
h1 = tf.nn.relu(h1)
with tf.variable_scope('h2'):
h2 = conv_pool2d(h1, nef * 2, kernel_size=5, strides=2)
h2 = norm_layer(h2)
h2 = tf.nn.relu(h2)
with tf.variable_scope('h3'):
h3 = conv_pool2d(h2, nef * 4, kernel_size=5, strides=2)
h3 = norm_layer(h3)
h3 = tf.nn.relu(h3)
h3_flatten = flatten(h3)
if include_top:
with tf.variable_scope('z_mu'):
z_mu = dense(h3_flatten, nz)
with tf.variable_scope('z_log_sigma_sq'):
z_log_sigma_sq = dense(h3_flatten, nz)
z_log_sigma_sq = tf.clip_by_value(z_log_sigma_sq, -10, 10)
outputs = {'enc_zs_mu': z_mu, 'enc_zs_log_sigma_sq': z_log_sigma_sq}
else:
outputs = h3_flatten
return outputs
def create_encoder(inputs,
e_net='legacy',
use_e_rnn=False,
rnn='lstm',
**kwargs):
assert inputs.shape.ndims == 5
batch_shape = inputs.shape[:-3].as_list()
inputs = flatten(inputs, 0, len(batch_shape) - 1)
unflatten = lambda x: tf.reshape(x, batch_shape + x.shape.as_list()[1:])
if use_e_rnn:
if e_net == 'legacy':
kwargs.pop('n_layers', None) # unused
h = create_legacy_encoder(inputs, include_top=False, **kwargs)
with tf.variable_scope('h4'):
h = dense(h, kwargs['nef'] * 4)
elif e_net == 'n_layer':
h = create_n_layer_encoder(inputs, include_top=False, **kwargs)
with tf.variable_scope('layer_%d' % (kwargs['n_layers'] + 1)):
h = dense(h, kwargs['nef'] * 4)
else:
raise ValueError('Invalid encoder net %s' % e_net)
if rnn == 'lstm':
RNNCell = tf.contrib.rnn.BasicLSTMCell
elif rnn == 'gru':
RNNCell = tf.contrib.rnn.GRUCell
else:
raise NotImplementedError
h = nest.map_structure(unflatten, h)
for i in range(2):
with tf.variable_scope('%s_h%d' % (rnn, i)):
rnn_cell = RNNCell(kwargs['nef'] * 4)
h, _ = tf.nn.dynamic_rnn(rnn_cell,
h,
dtype=tf.float32,
time_major=True)
h = flatten(h, 0, len(batch_shape) - 1)
with tf.variable_scope('z_mu'):
z_mu = dense(h, kwargs['nz'])
with tf.variable_scope('z_log_sigma_sq'):
z_log_sigma_sq = dense(h, kwargs['nz'])
z_log_sigma_sq = tf.clip_by_value(z_log_sigma_sq, -10, 10)
outputs = {'enc_zs_mu': z_mu, 'enc_zs_log_sigma_sq': z_log_sigma_sq}
else:
if e_net == 'legacy':
kwargs.pop('n_layers', None) # unused
outputs = create_legacy_encoder(inputs, include_top=True, **kwargs)
elif e_net == 'n_layer':
outputs = create_n_layer_encoder(inputs,
include_top=True,
**kwargs)
else:
raise ValueError('Invalid encoder net %s' % e_net)
outputs = nest.map_structure(unflatten, outputs)
return outputs
def create_n_layer_encoder(inputs,
nz=8,
nef=64,
n_layers=3,
norm_layer='instance',
include_top=True):
norm_layer = ops.get_norm_layer(norm_layer)
layers = []
paddings = [[0, 0], [1, 1], [1, 1], [0, 0]]
with tf.variable_scope("layer_1"):
convolved = conv2d(tf.pad(inputs, paddings),
nef,
kernel_size=4,
strides=2,
padding='VALID')
rectified = lrelu(convolved, 0.2)
layers.append(rectified)
for i in range(1, n_layers):
with tf.variable_scope("layer_%d" % (len(layers) + 1)):
out_channels = nef * min(2**i, 4)
convolved = conv2d(tf.pad(layers[-1], paddings),
out_channels,
kernel_size=4,
strides=2,
padding='VALID')
normalized = norm_layer(convolved)
rectified = lrelu(normalized, 0.2)
layers.append(rectified)
pooled = pool2d(rectified,
rectified.shape[1:3].as_list(),
padding='VALID',
pool_mode='avg')
squeezed = tf.squeeze(pooled, [1, 2])
if include_top:
with tf.variable_scope('z_mu'):
z_mu = dense(squeezed, nz)
with tf.variable_scope('z_log_sigma_sq'):
z_log_sigma_sq = dense(squeezed, nz)
z_log_sigma_sq = tf.clip_by_value(z_log_sigma_sq, -10, 10)
outputs = {'enc_zs_mu': z_mu, 'enc_zs_log_sigma_sq': z_log_sigma_sq}
else:
outputs = squeezed
return outputs
def video_sn_discriminator(clips, ndf=64):
clips = tf_utils.transpose_batch_time(clips)
batch_size = clips.shape[0].value
layers = []
paddings = [[0, 0], [1, 1], [1, 1], [1, 1], [0, 0]]
def conv3d(inputs, *args, **kwargs):
kwargs.setdefault('padding', 'VALID')
kwargs.setdefault('use_spectral_norm', True)
return ops.conv3d(tf.pad(inputs, paddings), *args, **kwargs)
with tf.variable_scope("sn_conv0_0"):
layers.append(lrelu(conv3d(clips, ndf, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_conv0_1"):
layers.append(
lrelu(
conv3d(layers[-1], ndf * 2, kernel_size=4, strides=(1, 2, 2)),
0.1))
with tf.variable_scope("sn_conv1_0"):
layers.append(
lrelu(conv3d(layers[-1], ndf * 2, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_conv1_1"):
layers.append(
lrelu(
conv3d(layers[-1], ndf * 4, kernel_size=4, strides=(1, 2, 2)),
0.1))
with tf.variable_scope("sn_conv2_0"):
layers.append(
lrelu(conv3d(layers[-1], ndf * 4, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_conv2_1"):
layers.append(
lrelu(conv3d(layers[-1], ndf * 8, kernel_size=4, strides=2), 0.1))
with tf.variable_scope("sn_conv3_0"):
layers.append(
lrelu(conv3d(layers[-1], ndf * 8, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_fc4"):
logits = dense(tf.reshape(layers[-1], [batch_size, -1]),
1,
use_spectral_norm=True)
layers.append(logits)
layers = nest.map_structure(tf_utils.transpose_batch_time, layers)
return layers
def image_sn_discriminator(images, ndf=64):
batch_size = images.shape[0].value
layers = []
paddings = [[0, 0], [1, 1], [1, 1], [0, 0]]
def conv2d(inputs, *args, **kwargs):
kwargs.setdefault('padding', 'VALID')
kwargs.setdefault('use_spectral_norm', True)
return ops.conv2d(tf.pad(inputs, paddings), *args, **kwargs)
with tf.variable_scope("sn_conv0_0"):
layers.append(lrelu(conv2d(images, ndf, kernel_size=3, strides=1),
0.1))
with tf.variable_scope("sn_conv0_1"):
layers.append(
lrelu(conv2d(layers[-1], ndf * 2, kernel_size=4, strides=2), 0.1))
with tf.variable_scope("sn_conv1_0"):
layers.append(
lrelu(conv2d(layers[-1], ndf * 2, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_conv1_1"):
layers.append(
lrelu(conv2d(layers[-1], ndf * 4, kernel_size=4, strides=2), 0.1))
with tf.variable_scope("sn_conv2_0"):
layers.append(
lrelu(conv2d(layers[-1], ndf * 4, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_conv2_1"):
layers.append(
lrelu(conv2d(layers[-1], ndf * 8, kernel_size=4, strides=2), 0.1))
with tf.variable_scope("sn_conv3_0"):
layers.append(
lrelu(conv2d(layers[-1], ndf * 8, kernel_size=3, strides=1), 0.1))
with tf.variable_scope("sn_fc4"):
logits = dense(tf.reshape(layers[-1], [batch_size, -1]),
1,
use_spectral_norm=True)
layers.append(logits)
return layers
def call(self, inputs, states):
norm_layer = ops.get_norm_layer(self.hparams.norm_layer)
downsample_layer = ops.get_downsample_layer(
self.hparams.downsample_layer)
upsample_layer = ops.get_upsample_layer(self.hparams.upsample_layer)
image_shape = inputs['images'].get_shape().as_list()
batch_size, height, width, color_channels = image_shape
time = states['time']
with tf.control_dependencies([tf.assert_equal(time[1:], time[0])]):
t = tf.to_int32(tf.identity(time[0]))
if 'states' in inputs:
state = tf.where(self.ground_truth[t], inputs['states'],
states['gen_state'])
state_action = []
state_action_z = []
if 'actions' in inputs:
state_action.append(inputs['actions'])
state_action_z.append(inputs['actions'])
if 'states' in inputs:
state_action.append(state)
# don't backpropagate the convnet through the state dynamics
state_action_z.append(tf.stop_gradient(state))
if 'zs' in inputs:
if self.hparams.use_rnn_z:
with tf.variable_scope('%s_z' % self.hparams.rnn):
rnn_z, rnn_z_state = self._rnn_func(
inputs['zs'], states['rnn_z_state'], self.hparams.nz)
state_action_z.append(rnn_z)
else:
state_action_z.append(inputs['zs'])
def concat(tensors, axis):
if len(tensors) == 0:
return tf.zeros([batch_size, 0])
elif len(tensors) == 1:
return tensors[0]
else:
return tf.concat(tensors, axis=axis)
state_action = concat(state_action, axis=-1)
state_action_z = concat(state_action_z, axis=-1)
image_views = []
first_image_views = []
if 'pix_distribs' in inputs:
pix_distrib_views = []
for i in range(self.hparams.num_views):
suffix = '%d' % i if i > 0 else ''
image_view = tf.where(
self.ground_truth[t], inputs['images' + suffix],
states['gen_image' + suffix]) # schedule sampling (if any)
image_views.append(image_view)
first_image_views.append(self.inputs['images' + suffix][0])
if 'pix_distribs' in inputs:
pix_distrib_view = tf.where(self.ground_truth[t],
inputs['pix_distribs' + suffix],
states['gen_pix_distrib' + suffix])
pix_distrib_views.append(pix_distrib_view)
outputs = {}
new_states = {}
all_layers = []
for i in range(self.hparams.num_views):
suffix = '%d' % i if i > 0 else ''
conv_rnn_states = states['conv_rnn_states' + suffix]
layers = []
new_conv_rnn_states = []
for i, (out_channels,
use_conv_rnn) in enumerate(self.encoder_layer_specs):
with tf.variable_scope('h%d' % i + suffix):
if i == 0:
# all image views and the first image corresponding to this view only
h = tf.concat(image_views + first_image_views, axis=-1)
kernel_size = (5, 5)
else:
h = layers[-1][-1]
kernel_size = (3, 3)
if self.hparams.where_add == 'all' or (
self.hparams.where_add == 'input' and i == 0):
h = tile_concat([h, state_action_z[:, None, None, :]],
axis=-1)
h = downsample_layer(h,
out_channels,
kernel_size=kernel_size,
strides=(2, 2))
h = norm_layer(h)
h = tf.nn.relu(h)
if use_conv_rnn:
conv_rnn_state = conv_rnn_states[len(new_conv_rnn_states)]
with tf.variable_scope('%s_h%d' %
(self.hparams.conv_rnn, i) +
suffix):
if self.hparams.where_add == 'all':
conv_rnn_h = tile_concat(
[h, state_action_z[:, None, None, :]], axis=-1)
else:
conv_rnn_h = h
conv_rnn_h, conv_rnn_state = self._conv_rnn_func(
conv_rnn_h, conv_rnn_state, out_channels)
new_conv_rnn_states.append(conv_rnn_state)
layers.append((h, conv_rnn_h) if use_conv_rnn else (h, ))
num_encoder_layers = len(layers)
for i, (out_channels,
use_conv_rnn) in enumerate(self.decoder_layer_specs):
with tf.variable_scope('h%d' % len(layers) + suffix):
if i == 0:
h = layers[-1][-1]
else:
h = tf.concat([
layers[-1][-1],
layers[num_encoder_layers - i - 1][-1]
],
axis=-1)
if self.hparams.where_add == 'all' or (
self.hparams.where_add == 'middle' and i == 0):
h = tile_concat([h, state_action_z[:, None, None, :]],
axis=-1)
h = upsample_layer(h,
out_channels,
kernel_size=(3, 3),
strides=(2, 2))
h = norm_layer(h)
h = tf.nn.relu(h)
if use_conv_rnn:
conv_rnn_state = conv_rnn_states[len(new_conv_rnn_states)]
with tf.variable_scope(
'%s_h%d' % (self.hparams.conv_rnn, len(layers)) +
suffix):
if self.hparams.where_add == 'all':
conv_rnn_h = tile_concat(
[h, state_action_z[:, None, None, :]], axis=-1)
else:
conv_rnn_h = h
conv_rnn_h, conv_rnn_state = self._conv_rnn_func(
conv_rnn_h, conv_rnn_state, out_channels)
new_conv_rnn_states.append(conv_rnn_state)
layers.append((h, conv_rnn_h) if use_conv_rnn else (h, ))
assert len(new_conv_rnn_states) == len(conv_rnn_states)
new_states['conv_rnn_states' + suffix] = new_conv_rnn_states
all_layers.append(layers)
if self.hparams.shared_views:
break
for i in range(self.hparams.num_views):
suffix = '%d' % i if i > 0 else ''
if self.hparams.shared_views:
layers, = all_layers
else:
layers = all_layers[i]
image = image_views[i]
last_images = states['last_images' + suffix][1:] + [image]
if 'pix_distribs' in inputs:
pix_distrib = pix_distrib_views[i]
last_pix_distribs = states['last_pix_distribs' +
suffix][1:] + [pix_distrib]
if self.hparams.last_frames and self.hparams.num_transformed_images:
if self.hparams.transformation == 'flow':
with tf.variable_scope('h%d_flow' % len(layers) + suffix):
h_flow = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_flow = norm_layer(h_flow)
h_flow = tf.nn.relu(h_flow)
with tf.variable_scope('flows' + suffix):
flows = conv2d(h_flow,
2 * self.hparams.last_frames *
self.hparams.num_transformed_images,
kernel_size=(3, 3),
strides=(1, 1))
flows = tf.reshape(flows, [
batch_size, height, width, 2,
self.hparams.last_frames *
self.hparams.num_transformed_images
])
else:
assert len(self.hparams.kernel_size) == 2
kernel_shape = list(self.hparams.kernel_size) + [
self.hparams.last_frames *
self.hparams.num_transformed_images
]
if self.hparams.transformation == 'dna':
with tf.variable_scope('h%d_dna_kernel' % len(layers) +
suffix):
h_dna_kernel = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_dna_kernel = norm_layer(h_dna_kernel)
h_dna_kernel = tf.nn.relu(h_dna_kernel)
# Using largest hidden state for predicting untied conv kernels.
with tf.variable_scope('dna_kernels' + suffix):
kernels = conv2d(h_dna_kernel,
np.prod(kernel_shape),
kernel_size=(3, 3),
strides=(1, 1))
kernels = tf.reshape(kernels,
[batch_size, height, width] +
kernel_shape)
kernels = kernels + identity_kernel(
self.hparams.kernel_size)[None, None,
None, :, :, None]
kernel_spatial_axes = [3, 4]
elif self.hparams.transformation == 'cdna':
with tf.variable_scope('cdna_kernels' + suffix):
smallest_layer = layers[num_encoder_layers - 1][-1]
kernels = dense(flatten(smallest_layer),
np.prod(kernel_shape))
kernels = tf.reshape(kernels,
[batch_size] + kernel_shape)
kernels = kernels + identity_kernel(
self.hparams.kernel_size)[None, :, :, None]
kernel_spatial_axes = [1, 2]
else:
raise ValueError('Invalid transformation %s' %
self.hparams.transformation)
if self.hparams.transformation != 'flow':
with tf.name_scope('kernel_normalization' + suffix):
kernels = tf.nn.relu(kernels - RELU_SHIFT) + RELU_SHIFT
kernels /= tf.reduce_sum(kernels,
axis=kernel_spatial_axes,
keepdims=True)
if self.hparams.generate_scratch_image:
with tf.variable_scope('h%d_scratch' % len(layers) + suffix):
h_scratch = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_scratch = norm_layer(h_scratch)
h_scratch = tf.nn.relu(h_scratch)
# Using largest hidden state for predicting a new image layer.
# This allows the network to also generate one image from scratch,
# which is useful when regions of the image become unoccluded.
with tf.variable_scope('scratch_image' + suffix):
scratch_image = conv2d(h_scratch,
color_channels,
kernel_size=(3, 3),
strides=(1, 1))
scratch_image = tf.nn.sigmoid(scratch_image)
with tf.name_scope('transformed_images' + suffix):
transformed_images = []
if self.hparams.last_frames and self.hparams.num_transformed_images:
if self.hparams.transformation == 'flow':
transformed_images.extend(
apply_flows(last_images, flows))
else:
transformed_images.extend(
apply_kernels(last_images, kernels,
self.hparams.dilation_rate))
if self.hparams.prev_image_background:
transformed_images.append(image)
if self.hparams.first_image_background and not self.hparams.context_images_background:
transformed_images.append(self.inputs['images' +
suffix][0])
if self.hparams.context_images_background:
transformed_images.extend(
tf.unstack(
self.inputs['images' +
suffix][:self.hparams.context_frames]))
if self.hparams.generate_scratch_image:
transformed_images.append(scratch_image)
if 'pix_distribs' in inputs:
with tf.name_scope('transformed_pix_distribs' + suffix):
transformed_pix_distribs = []
if self.hparams.last_frames and self.hparams.num_transformed_images:
if self.hparams.transformation == 'flow':
transformed_pix_distribs.extend(
apply_flows(last_pix_distribs, flows))
else:
transformed_pix_distribs.extend(
apply_kernels(last_pix_distribs, kernels,
self.hparams.dilation_rate))
if self.hparams.prev_image_background:
transformed_pix_distribs.append(pix_distrib)
if self.hparams.first_image_background and not self.hparams.context_images_background:
transformed_pix_distribs.append(
self.inputs['pix_distribs' + suffix][0])
if self.hparams.context_images_background:
transformed_pix_distribs.extend(
tf.unstack(self.inputs['pix_distribs' + suffix]
[:self.hparams.context_frames]))
if self.hparams.generate_scratch_image:
transformed_pix_distribs.append(pix_distrib)
with tf.name_scope('masks' + suffix):
if len(transformed_images) > 1:
with tf.variable_scope('h%d_masks' % len(layers) + suffix):
h_masks = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_masks = norm_layer(h_masks)
h_masks = tf.nn.relu(h_masks)
with tf.variable_scope('masks' + suffix):
if self.hparams.dependent_mask:
h_masks = tf.concat([h_masks] + transformed_images,
axis=-1)
masks = conv2d(h_masks,
len(transformed_images),
kernel_size=(3, 3),
strides=(1, 1))
masks = tf.nn.softmax(masks)
masks = tf.split(masks,
len(transformed_images),
axis=-1)
elif len(transformed_images) == 1:
masks = [tf.ones([batch_size, height, width, 1])]
else:
raise ValueError(
"Either one of the following should be true: "
"last_frames and num_transformed_images, first_image_background, "
"prev_image_background, generate_scratch_image")
with tf.name_scope('gen_images' + suffix):
assert len(transformed_images) == len(masks)
gen_image = tf.add_n([
transformed_image * mask for transformed_image, mask in
zip(transformed_images, masks)
])
if 'pix_distribs' in inputs:
with tf.name_scope('gen_pix_distribs' + suffix):
assert len(transformed_pix_distribs) == len(masks)
gen_pix_distrib = tf.add_n([
transformed_pix_distrib * mask
for transformed_pix_distrib, mask in zip(
transformed_pix_distribs, masks)
])
if self.hparams.renormalize_pixdistrib:
gen_pix_distrib /= tf.reduce_sum(gen_pix_distrib,
axis=(1, 2),
keepdims=True)
outputs['gen_images' + suffix] = gen_image
outputs['transformed_images' + suffix] = tf.stack(
transformed_images, axis=-1)
outputs['masks' + suffix] = tf.stack(masks, axis=-1)
if 'pix_distribs' in inputs:
outputs['gen_pix_distribs' + suffix] = gen_pix_distrib
outputs['transformed_pix_distribs' + suffix] = tf.stack(
transformed_pix_distribs, axis=-1)
if self.hparams.transformation == 'flow':
outputs['gen_flows' + suffix] = flows
flows_transposed = tf.transpose(flows, [0, 1, 2, 4, 3])
flows_rgb_transposed = tf_utils.flow_to_rgb(flows_transposed)
flows_rgb = tf.transpose(flows_rgb_transposed, [0, 1, 2, 4, 3])
outputs['gen_flows_rgb' + suffix] = flows_rgb
new_states['gen_image' + suffix] = gen_image
new_states['last_images' + suffix] = last_images
if 'pix_distribs' in inputs:
new_states['gen_pix_distrib' + suffix] = gen_pix_distrib
new_states['last_pix_distribs' + suffix] = last_pix_distribs
if 'states' in inputs:
with tf.name_scope('gen_states'):
with tf.variable_scope('state_pred'):
gen_state = dense(state_action,
inputs['states'].shape[-1].value)
if 'states' in inputs:
outputs['gen_states'] = gen_state
new_states['time'] = time + 1
if 'zs' in inputs and self.hparams.use_rnn_z:
new_states['rnn_z_state'] = rnn_z_state
if 'states' in inputs:
new_states['gen_state'] = gen_state
return outputs, new_states
def call(self, inputs, states):
norm_layer = ops.get_norm_layer(self.hparams.norm_layer)
image_shape = inputs['images'].get_shape().as_list()
batch_size, height, width, color_channels = image_shape
conv_rnn_states = states['conv_rnn_states']
time = states['time']
with tf.control_dependencies([tf.assert_equal(time[1:], time[0])]):
t = tf.to_int32(tf.identity(time[0]))
last_gt_images = states['last_gt_images'][1:] + [inputs['images']]
last_pred_flows = states['last_pred_flows'][1:] + [
tf.zeros_like(states['last_pred_flows'][-1])
]
image = tf.where(self.ground_truth[t], inputs['images'],
states['gen_image'])
last_images = states['last_images'][1:] + [image]
last_base_images = [
tf.where(self.ground_truth[t], last_gt_image, last_base_image)
for last_gt_image, last_base_image in zip(
last_gt_images, states['last_base_images'])
]
last_base_flows = [
tf.where(self.ground_truth[t], last_pred_flow, last_base_flow)
for last_pred_flow, last_base_flow in zip(
last_pred_flows, states['last_base_flows'])
]
if 'pix_distribs' in inputs:
last_gt_pix_distribs = states['last_gt_pix_distribs'][1:] + [
inputs['pix_distribs']
]
last_base_pix_distribs = [
tf.where(self.ground_truth[t], last_gt_pix_distrib,
last_base_pix_distrib)
for last_gt_pix_distrib, last_base_pix_distrib in zip(
last_gt_pix_distribs, states['last_base_pix_distribs'])
]
if 'states' in inputs:
state = tf.where(self.ground_truth[t], inputs['states'],
states['gen_state'])
state_action = []
state_action_z = []
if 'actions' in inputs:
state_action.append(inputs['actions'])
state_action_z.append(inputs['actions'])
if 'states' in inputs:
state_action.append(state)
# don't backpropagate the convnet through the state dynamics
state_action_z.append(tf.stop_gradient(state))
if 'zs' in inputs:
if self.hparams.use_rnn_z:
with tf.variable_scope('%s_z' % self.hparams.rnn):
rnn_z, rnn_z_state = self._rnn_func(
inputs['zs'], states['rnn_z_state'], self.hparams.nz)
state_action_z.append(rnn_z)
else:
state_action_z.append(inputs['zs'])
def concat(tensors, axis):
if len(tensors) == 0:
return tf.zeros([batch_size, 0])
elif len(tensors) == 1:
return tensors[0]
else:
return tf.concat(tensors, axis=axis)
state_action = concat(state_action, axis=-1)
state_action_z = concat(state_action_z, axis=-1)
if 'actions' in inputs:
gen_input = tile_concat(last_images +
[inputs['actions'][:, None, None, :]],
axis=-1)
else:
gen_input = tf.concat(last_images)
layers = []
new_conv_rnn_states = []
for i, (out_channels,
use_conv_rnn) in enumerate(self.encoder_layer_specs):
with tf.variable_scope('h%d' % i):
if i == 0:
h = tf.concat(last_images, axis=-1)
kernel_size = (5, 5)
else:
h = layers[-1][-1]
kernel_size = (3, 3)
h = conv_pool2d(tile_concat(
[h, state_action_z[:, None, None, :]], axis=-1),
out_channels,
kernel_size=kernel_size,
strides=(2, 2))
h = norm_layer(h)
h = tf.nn.relu(h)
if use_conv_rnn:
conv_rnn_state = conv_rnn_states[len(new_conv_rnn_states)]
with tf.variable_scope('%s_h%d' % (self.hparams.conv_rnn, i)):
conv_rnn_h, conv_rnn_state = self._conv_rnn_func(
tile_concat([h, state_action_z[:, None, None, :]],
axis=-1), conv_rnn_state, out_channels)
new_conv_rnn_states.append(conv_rnn_state)
layers.append((h, conv_rnn_h) if use_conv_rnn else (h, ))
num_encoder_layers = len(layers)
for i, (out_channels,
use_conv_rnn) in enumerate(self.decoder_layer_specs):
with tf.variable_scope('h%d' % len(layers)):
if i == 0:
h = layers[-1][-1]
else:
h = tf.concat([
layers[-1][-1], layers[num_encoder_layers - i - 1][-1]
],
axis=-1)
h = upsample_conv2d(tile_concat(
[h, state_action_z[:, None, None, :]], axis=-1),
out_channels,
kernel_size=(3, 3),
strides=(2, 2))
h = norm_layer(h)
h = tf.nn.relu(h)
if use_conv_rnn:
conv_rnn_state = conv_rnn_states[len(new_conv_rnn_states)]
with tf.variable_scope('%s_h%d' %
(self.hparams.conv_rnn, len(layers))):
conv_rnn_h, conv_rnn_state = self._conv_rnn_func(
tile_concat([h, state_action_z[:, None, None, :]],
axis=-1), conv_rnn_state, out_channels)
new_conv_rnn_states.append(conv_rnn_state)
layers.append((h, conv_rnn_h) if use_conv_rnn else (h, ))
assert len(new_conv_rnn_states) == len(conv_rnn_states)
with tf.variable_scope('h%d_flow' % len(layers)):
h_flow = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_flow = norm_layer(h_flow)
h_flow = tf.nn.relu(h_flow)
with tf.variable_scope('flows'):
flows = conv2d(h_flow,
2 * self.hparams.last_frames,
kernel_size=(3, 3),
strides=(1, 1))
flows = tf.reshape(
flows,
[batch_size, height, width, 2, self.hparams.last_frames])
with tf.name_scope('transformed_images'):
transformed_images = []
last_pred_flows = [
flow + flow_ops.image_warp(last_pred_flow, flow)
for last_pred_flow, flow in zip(last_pred_flows,
tf.unstack(flows, axis=-1))
]
last_base_flows = [
flow + flow_ops.image_warp(last_base_flow, flow)
for last_base_flow, flow in zip(last_base_flows,
tf.unstack(flows, axis=-1))
]
for last_base_image, last_base_flow in zip(last_base_images,
last_base_flows):
transformed_images.append(
flow_ops.image_warp(last_base_image, last_base_flow))
if 'pix_distribs' in inputs:
with tf.name_scope('transformed_pix_distribs'):
transformed_pix_distribs = []
for last_base_pix_distrib, last_base_flow in zip(
last_base_pix_distribs, last_base_flows):
transformed_pix_distribs.append(
flow_ops.image_warp(last_base_pix_distrib,
last_base_flow))
with tf.name_scope('masks'):
if len(transformed_images) > 1:
with tf.variable_scope('h%d_masks' % len(layers)):
h_masks = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_masks = norm_layer(h_masks)
h_masks = tf.nn.relu(h_masks)
with tf.variable_scope('masks'):
if self.hparams.dependent_mask:
h_masks = tf.concat([h_masks] + transformed_images,
axis=-1)
masks = conv2d(h_masks,
len(transformed_images),
kernel_size=(3, 3),
strides=(1, 1))
masks = tf.nn.softmax(masks)
masks = tf.split(masks, len(transformed_images), axis=-1)
elif len(transformed_images) == 1:
masks = [tf.ones([batch_size, height, width, 1])]
else:
raise ValueError(
"Either one of the following should be true: "
"last_frames and num_transformed_images, first_image_background, "
"prev_image_background, generate_scratch_image")
with tf.name_scope('gen_images'):
assert len(transformed_images) == len(masks)
gen_image = tf.add_n([
transformed_image * mask
for transformed_image, mask in zip(transformed_images, masks)
])
if 'pix_distribs' in inputs:
with tf.name_scope('gen_pix_distribs'):
assert len(transformed_pix_distribs) == len(masks)
gen_pix_distrib = tf.add_n([
transformed_pix_distrib * mask
for transformed_pix_distrib, mask in zip(
transformed_pix_distribs, masks)
])
# TODO: is this needed?
# gen_pix_distrib /= tf.reduce_sum(gen_pix_distrib, axis=(1, 2), keepdims=True)
if 'states' in inputs:
with tf.name_scope('gen_states'):
with tf.variable_scope('state_pred'):
gen_state = dense(state_action,
inputs['states'].shape[-1].value)
outputs = {
'gen_images': gen_image,
'gen_inputs': gen_input,
'transformed_images': tf.stack(transformed_images, axis=-1),
'masks': tf.stack(masks, axis=-1),
'gen_flow': flows
}
if 'pix_distribs' in inputs:
outputs['gen_pix_distribs'] = gen_pix_distrib
outputs['transformed_pix_distribs'] = tf.stack(
transformed_pix_distribs, axis=-1)
if 'states' in inputs:
outputs['gen_states'] = gen_state
new_states = {
'time': time + 1,
'last_gt_images': last_gt_images,
'last_pred_flows': last_pred_flows,
'gen_image': gen_image,
'last_images': last_images,
'last_base_images': last_base_images,
'last_base_flows': last_base_flows,
'conv_rnn_states': new_conv_rnn_states
}
if 'zs' in inputs and self.hparams.use_rnn_z:
new_states['rnn_z_state'] = rnn_z_state
if 'pix_distribs' in inputs:
new_states['last_gt_pix_distribs'] = last_gt_pix_distribs
new_states['last_base_pix_distribs'] = last_base_pix_distribs
if 'states' in inputs:
new_states['gen_state'] = gen_state
return outputs, new_states
def generator_fn(inputs, mode, hparams=None):
inputs = {name: tf_utils.maybe_pad_or_slice(input, hparams.sequence_length - 1)
for name, input in inputs.items()}
images = inputs['images']
input_images = tf.concat(tf.unstack(images[:hparams.context_frames], axis=0), axis=-1)
gen_states = []
for t in range(hparams.context_frames, hparams.sequence_length):
state_action = []
if 'actions' in inputs:
state_action.append(inputs['actions'][t - 1])
if 'states' in inputs:
state_action.append(gen_states[-1] if gen_states else inputs['states'][t - 1])
state_action = tf.concat(state_action, axis=-1)
with tf.name_scope('gen_states'):
with tf.variable_scope('state_pred%d' % t):
gen_state = dense(state_action, inputs['states'].shape[-1])
gen_states.append(gen_state)
states_actions = []
if 'actions' in inputs:
states_actions += tf.unstack(inputs['actions'][:hparams.sequence_length - 1], axis=0)
if 'states' in inputs:
states_actions += tf.unstack(inputs['states'][:hparams.context_frames], axis=0)
states_actions += gen_states
if states_actions:
states_actions = tf.concat(states_actions, axis=-1)
# don't backpropagate the convnet through the state dynamics
states_actions = tf.stop_gradient(states_actions)
else:
states_actions = tf.zeros([images.shape[1], 0])
with slim.arg_scope([slim.conv2d],
activation_fn=tf.nn.relu,
weights_initializer=tf.truncated_normal_initializer(0.0, 0.01),
weights_regularizer=slim.l2_regularizer(0.0001)):
batch_norm_params = {
'decay': 0.9997,
'epsilon': 0.001,
'is_training': mode == 'train',
}
with slim.arg_scope([slim.batch_norm], is_training=mode == 'train', updates_collections=None):
with slim.arg_scope([slim.conv2d], normalizer_fn=slim.batch_norm,
normalizer_params=batch_norm_params):
h0 = slim.conv2d(input_images, 64, [5, 5], stride=1, scope='conv1')
size0 = tf.shape(input_images)[-3:-1]
h1 = slim.max_pool2d(h0, [2, 2], scope='pool1')
h1 = slim.conv2d(h1, 128, [5, 5], stride=1, scope='conv2')
size1 = tf.shape(h1)[-3:-1]
h2 = slim.max_pool2d(h1, [2, 2], scope='pool2')
h2 = slim.conv2d(h2, 256, [3, 3], stride=1, scope='conv3')
size2 = tf.shape(h2)[-3:-1]
h3 = slim.max_pool2d(h2, [2, 2], scope='pool3')
h3 = tile_concat([h3, states_actions[:, None, None, :]], axis=-1)
h3 = slim.conv2d(h3, 256, [3, 3], stride=1, scope='conv4')
h4 = tf.image.resize_bilinear(h3, size2)
h4 = tf.concat([h4, h2], axis=-1)
h4 = slim.conv2d(h4, 256, [3, 3], stride=1, scope='conv5')
h5 = tf.image.resize_bilinear(h4, size1)
h5 = tf.concat([h5, h1], axis=-1)
h5 = slim.conv2d(h5, 128, [5, 5], stride=1, scope='conv6')
h6 = tf.image.resize_bilinear(h5, size0)
h6 = tf.concat([h6, h0], axis=-1)
h6 = slim.conv2d(h6, 64, [5, 5], stride=1, scope='conv7')
extrap_length = hparams.sequence_length - hparams.context_frames
flows_masks = slim.conv2d(h6, 5 * extrap_length, [5, 5], stride=1, activation_fn=tf.tanh,
normalizer_fn=None, scope='conv8')
flows_masks = tf.split(flows_masks, extrap_length, axis=-1)
gen_images = []
gen_flows_1 = []
gen_flows_2 = []
masks = []
for flows_mask in flows_masks:
flow_1, flow_2, mask = tf.split(flows_mask, [2, 2, 1], axis=-1)
gen_flows_1.append(flow_1)
gen_flows_2.append(flow_2)
mask = 0.5 * (1.0 + mask)
masks.append(mask)
linspace_x = tf.linspace(-1.0, 1.0, size0[1])
linspace_x.set_shape(input_images.shape[-2])
linspace_y = tf.linspace(-1.0, 1.0, size0[0])
linspace_y.set_shape(input_images.shape[-3])
grid_x, grid_y = tf.meshgrid(linspace_x, linspace_y)
coor_x_1 = grid_x[None, :, :] + flow_1[:, :, :, 0]
coor_y_1 = grid_y[None, :, :] + flow_1[:, :, :, 1]
coor_x_2 = grid_x[None, :, :] + flow_2[:, :, :, 0]
coor_y_2 = grid_y[None, :, :] + flow_2[:, :, :, 1]
output_1 = bilinear_interp(images[0], coor_x_1, coor_y_1, 'interpolate')
output_2 = bilinear_interp(images[1], coor_x_2, coor_y_2, 'interpolate')
gen_image = mask * output_1 + (1.0 - mask) * output_2
gen_images.append(gen_image)
gen_images = tf.stack(gen_images, axis=0)
gen_flows_1 = tf.stack(gen_flows_1, axis=0)
gen_flows_2 = tf.stack(gen_flows_2, axis=0)
masks = tf.stack(masks, axis=0)
outputs = {
'gen_images': gen_images,
'gen_flows_1': gen_flows_1,
'gen_flows_2': gen_flows_2,
'masks': masks,
}
if 'states' in inputs:
gen_states = tf.stack(gen_states, axis=0)
outputs['gen_states'] = gen_states
return gen_images, outputs
def call(self, inputs, states):
norm_layer = ops.get_norm_layer(self.hparams.norm_layer)
feature_shape = inputs['features'].get_shape().as_list()
batch_size, height, width, feature_channels = feature_shape
conv_rnn_states = states['conv_rnn_states']
time = states['time']
with tf.control_dependencies([tf.assert_equal(time[1:], time[0])]):
t = tf.to_int32(tf.identity(time[0]))
feature = tf.where(self.ground_truth[t], inputs['features'],
states['gen_feature']) # schedule sampling (if any)
if 'states' in inputs:
state = tf.where(self.ground_truth[t], inputs['states'],
states['gen_state'])
state_action = []
state_action_z = []
if 'actions' in inputs:
state_action.append(inputs['actions'])
state_action_z.append(inputs['actions'])
if 'states' in inputs:
state_action.append(state)
# don't backpropagate the convnet through the state dynamics
state_action_z.append(tf.stop_gradient(state))
if 'zs' in inputs:
if self.hparams.use_rnn_z:
with tf.variable_scope('%s_z' % self.hparams.rnn):
rnn_z, rnn_z_state = self._rnn_func(
inputs['zs'], states['rnn_z_state'], self.hparams.nz)
state_action_z.append(rnn_z)
else:
state_action_z.append(inputs['zs'])
def concat(tensors, axis):
if len(tensors) == 0:
return tf.zeros([batch_size, 0])
elif len(tensors) == 1:
return tensors[0]
else:
return tf.concat(tensors, axis=axis)
state_action = concat(state_action, axis=-1)
state_action_z = concat(state_action_z, axis=-1)
if 'actions' in inputs:
gen_input = tile_concat(
[feature, inputs['actions'][:, None, None, :]], axis=-1)
else:
gen_input = feature
layers = []
new_conv_rnn_states = []
for i, (out_channels,
use_conv_rnn) in enumerate(self.encoder_layer_specs):
with tf.variable_scope('h%d' % i):
if i == 0:
# h = tf.concat([feature, self.inputs['features'][0]], axis=-1) # TODO: use first feature?
h = feature
else:
h = layers[-1][-1]
h = conv_pool2d(tile_concat(
[h, state_action_z[:, None, None, :]], axis=-1),
out_channels,
kernel_size=(3, 3),
strides=(2, 2))
h = norm_layer(h)
h = tf.nn.relu(h)
if use_conv_rnn:
conv_rnn_state = conv_rnn_states[len(new_conv_rnn_states)]
with tf.variable_scope('%s_h%d' % (self.hparams.conv_rnn, i)):
conv_rnn_h, conv_rnn_state = self._conv_rnn_func(
tile_concat([h, state_action_z[:, None, None, :]],
axis=-1), conv_rnn_state, out_channels)
new_conv_rnn_states.append(conv_rnn_state)
layers.append((h, conv_rnn_h) if use_conv_rnn else (h, ))
num_encoder_layers = len(layers)
for i, (out_channels,
use_conv_rnn) in enumerate(self.decoder_layer_specs):
with tf.variable_scope('h%d' % len(layers)):
if i == 0:
h = layers[-1][-1]
else:
h = tf.concat([
layers[-1][-1], layers[num_encoder_layers - i - 1][-1]
],
axis=-1)
h = upsample_conv2d(tile_concat(
[h, state_action_z[:, None, None, :]], axis=-1),
out_channels,
kernel_size=(3, 3),
strides=(2, 2))
h = norm_layer(h)
h = tf.nn.relu(h)
if use_conv_rnn:
conv_rnn_state = conv_rnn_states[len(new_conv_rnn_states)]
with tf.variable_scope('%s_h%d' %
(self.hparams.conv_rnn, len(layers))):
conv_rnn_h, conv_rnn_state = self._conv_rnn_func(
tile_concat([h, state_action_z[:, None, None, :]],
axis=-1), conv_rnn_state, out_channels)
new_conv_rnn_states.append(conv_rnn_state)
layers.append((h, conv_rnn_h) if use_conv_rnn else (h, ))
assert len(new_conv_rnn_states) == len(conv_rnn_states)
if self.hparams.transformation == 'direct':
with tf.variable_scope('h%d_direct' % len(layers)):
h_direct = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_direct = norm_layer(h_direct)
h_direct = tf.nn.relu(h_direct)
with tf.variable_scope('direct'):
gen_feature = conv2d(h_direct,
feature_channels,
kernel_size=(3, 3),
strides=(1, 1))
else:
if self.hparams.transformation == 'flow':
with tf.variable_scope('h%d_flow' % len(layers)):
h_flow = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_flow = norm_layer(h_flow)
h_flow = tf.nn.relu(h_flow)
with tf.variable_scope('flows'):
flows = conv2d(h_flow,
2 * feature_channels,
kernel_size=(3, 3),
strides=(1, 1))
flows = tf.reshape(
flows,
[batch_size, height, width, 2, feature_channels])
transformations = flows
else:
assert len(self.hparams.kernel_size) == 2
kernel_shape = list(
self.hparams.kernel_size) + [feature_channels]
if self.hparams.transformation == 'local':
with tf.variable_scope('h%d_local_kernel' % len(layers)):
h_local_kernel = conv2d(layers[-1][-1],
self.hparams.ngf,
kernel_size=(3, 3),
strides=(1, 1))
h_local_kernel = norm_layer(h_local_kernel)
h_local_kernel = tf.nn.relu(h_local_kernel)
# Using largest hidden state for predicting untied conv kernels.
with tf.variable_scope('local_kernels'):
kernels = conv2d(h_local_kernel,
np.prod(kernel_shape),
kernel_size=(3, 3),
strides=(1, 1))
kernels = tf.reshape(kernels,
[batch_size, height, width] +
kernel_shape)
kernels = kernels + identity_kernel(
self.hparams.kernel_size)[None, None, None, :, :,
None]
elif self.hparams.transformation == 'conv':
with tf.variable_scope('conv_kernels'):
smallest_layer = layers[num_encoder_layers - 1][-1]
kernels = dense(flatten(smallest_layer),
np.prod(kernel_shape))
kernels = tf.reshape(kernels,
[batch_size] + kernel_shape)
kernels = kernels + identity_kernel(
self.hparams.kernel_size)[None, :, :, None]
else:
raise ValueError('Invalid transformation %s' %
self.hparams.transformation)
transformations = kernels
with tf.name_scope('gen_features'):
if self.hparams.transformation == 'flow':
def apply_transformation(feature_and_flow):
feature, flow = feature_and_flow
return flow_ops.image_warp(feature[..., None], flow)
else:
def apply_transformation(feature_and_kernel):
feature, kernel = feature_and_kernel
output, = apply_kernels(feature[..., None],
kernel[..., None])
return tf.squeeze(output, axis=-1)
gen_feature_transposed = tf.map_fn(
apply_transformation,
(tf.stack(tf.unstack(feature, axis=-1)),
tf.stack(tf.unstack(transformations, axis=-1))),
dtype=tf.float32)
gen_feature = tf.stack(tf.unstack(gen_feature_transposed),
axis=-1)
# TODO: use norm and relu for generated features?
gen_feature = norm_layer(gen_feature)
gen_feature = tf.nn.relu(gen_feature)
if 'states' in inputs:
with tf.name_scope('gen_states'):
with tf.variable_scope('state_pred'):
gen_state = dense(state_action,
inputs['states'].shape[-1].value)
outputs = {
'gen_features': gen_feature,
'gen_inputs': gen_input,
}
if 'states' in inputs:
outputs['gen_states'] = gen_state
if self.hparams.transformation == 'flow':
outputs['gen_flows'] = flows
new_states = {
'time': time + 1,
'gen_feature': gen_feature,
'conv_rnn_states': new_conv_rnn_states,
}
if 'zs' in inputs and self.hparams.use_rnn_z:
new_states['rnn_z_state'] = rnn_z_state
if 'states' in inputs:
new_states['gen_state'] = gen_state
return outputs, new_states
