def main(unused_argv):
tf.logging.set_verbosity(tf.logging.INFO)
dataset = data_generator.Dataset(
dataset_name=FLAGS.dataset,
split_name=FLAGS.eval_split,
dataset_dir=FLAGS.dataset_dir,
batch_size=FLAGS.eval_batch_size,
crop_size=[int(sz) for sz in FLAGS.eval_crop_size],
min_resize_value=FLAGS.min_resize_value,
max_resize_value=FLAGS.max_resize_value,
resize_factor=FLAGS.resize_factor,
model_variant=FLAGS.model_variant,
num_readers=2,
is_training=False,
should_shuffle=False,
should_repeat=False,
with_cls=True,
cls_only=False,
output_valid=True)
tf.gfile.MakeDirs(FLAGS.eval_logdir)
tf.logging.info('Evaluating on %s set', FLAGS.eval_split)
with tf.Graph().as_default():
samples = dataset.get_one_shot_iterator().get_next()
model_options = common.ModelOptions(
outputs_to_num_classes={
common.OUTPUT_TYPE: dataset.num_of_classes
},
crop_size=[int(sz) for sz in FLAGS.eval_crop_size],
atrous_rates=FLAGS.atrous_rates,
output_stride=FLAGS.output_stride)
# Set shape in order for tf.contrib.tfprof.model_analyzer to work properly.
samples[common.IMAGE].set_shape([
FLAGS.eval_batch_size,
int(FLAGS.eval_crop_size[0]),
int(FLAGS.eval_crop_size[1]), 3
])
if tuple(FLAGS.eval_scales) == (1.0, ):
tf.logging.info('Performing single-scale test.')
predictions = model.predict_labels(
samples[common.IMAGE],
model_options,
image_pyramid=FLAGS.image_pyramid)
else:
tf.logging.info('Performing multi-scale test.')
raise NotImplementedError('Multi-scale is not supported yet!')
metric_map = {}
## Extract cls logits
if FLAGS.weakly:
_, end_points = feature_extractor.extract_features(
samples[common.IMAGE],
output_stride=model_options.output_stride,
multi_grid=model_options.multi_grid,
model_variant=model_options.model_variant,
depth_multiplier=model_options.depth_multiplier,
divisible_by=model_options.divisible_by,
reuse=tf.AUTO_REUSE,
is_training=False,
preprocessed_images_dtype=model_options.
preprocessed_images_dtype,
global_pool=True,
num_classes=dataset.num_of_classes - 1)
# ResNet beta version has an additional suffix in FLAGS.model_variant, but
# it shares the same variable names with original version. Add a special
# handling here for beta version ResNet.
logits = end_points['{}/logits'.format(
FLAGS.model_variant).replace('_beta', '')]
logits = tf.reshape(logits, [-1, dataset.num_of_classes - 1])
cls_pred = tf.sigmoid(logits)
# Multi-label classification evaluation
cls_label = samples['cls_label']
cls_pred = tf.cast(tf.greater_equal(cls_pred, 0.5), tf.int32)
## For classification
metric_map['eval/cls_overall'] = tf.metrics.accuracy(
labels=cls_label, predictions=cls_pred)
metric_map['eval/cls_precision'] = tf.metrics.precision(
labels=cls_label, predictions=cls_pred)
metric_map['eval/cls_recall'] = tf.metrics.recall(
labels=cls_label, predictions=cls_pred)
## For segmentation branch eval
predictions = predictions[common.OUTPUT_TYPE]
predictions = tf.reshape(predictions, shape=[-1])
labels = tf.reshape(samples[common.LABEL], shape=[-1])
weights = tf.to_float(tf.not_equal(labels, dataset.ignore_label))
# Set ignore_label regions to label 0, because metrics.mean_iou requires
# range of labels = [0, dataset.num_classes). Note the ignore_label regions
# are not evaluated since the corresponding regions contain weights = 0.
labels = tf.where(tf.equal(labels, dataset.ignore_label),
tf.zeros_like(labels), labels)
predictions_tag = 'miou'
# Define the evaluation metric.
num_classes = dataset.num_of_classes
## For segmentation
metric_map['eval/%s_overall' % predictions_tag] = tf.metrics.mean_iou(
labels=labels,
predictions=predictions,
num_classes=num_classes,
weights=weights)
# IoU for each class.
one_hot_predictions = tf.one_hot(predictions, num_classes)
one_hot_predictions = tf.reshape(one_hot_predictions,
[-1, num_classes])
one_hot_labels = tf.one_hot(labels, num_classes)
one_hot_labels = tf.reshape(one_hot_labels, [-1, num_classes])
for c in range(num_classes):
predictions_tag_c = '%s_class_%d' % (predictions_tag, c)
tp, tp_op = tf.metrics.true_positives(
labels=one_hot_labels[:, c],
predictions=one_hot_predictions[:, c],
weights=weights)
fp, fp_op = tf.metrics.false_positives(
labels=one_hot_labels[:, c],
predictions=one_hot_predictions[:, c],
weights=weights)
fn, fn_op = tf.metrics.false_negatives(
labels=one_hot_labels[:, c],
predictions=one_hot_predictions[:, c],
weights=weights)
tp_fp_fn_op = tf.group(tp_op, fp_op, fn_op)
iou = tf.where(tf.greater(tp + fn, 0.0), tp / (tp + fn + fp),
tf.constant(np.NaN))
metric_map['eval/%s' % predictions_tag_c] = (iou, tp_fp_fn_op)
(metrics_to_values,
metrics_to_updates) = contrib_metrics.aggregate_metric_map(metric_map)
summary_ops = []
for metric_name, metric_value in six.iteritems(metrics_to_values):
op = tf.summary.scalar(metric_name, metric_value)
op = tf.Print(op, [metric_value], metric_name)
summary_ops.append(op)
summary_op = tf.summary.merge(summary_ops)
summary_hook = contrib_training.SummaryAtEndHook(
log_dir=FLAGS.eval_logdir, summary_op=summary_op)
hooks = [summary_hook]
num_eval_iters = None
if FLAGS.max_number_of_evaluations > 0:
num_eval_iters = FLAGS.max_number_of_evaluations
if FLAGS.quantize_delay_step >= 0:
contrib_quantize.create_eval_graph()
contrib_tfprof.model_analyzer.print_model_analysis(
tf.get_default_graph(),
tfprof_options=contrib_tfprof.model_analyzer.
TRAINABLE_VARS_PARAMS_STAT_OPTIONS)
contrib_tfprof.model_analyzer.print_model_analysis(
tf.get_default_graph(),
tfprof_options=contrib_tfprof.model_analyzer.FLOAT_OPS_OPTIONS)
contrib_training.evaluate_repeatedly(
checkpoint_dir=FLAGS.checkpoint_dir,
master=FLAGS.master,
eval_ops=list(metrics_to_updates.values()),
max_number_of_evaluations=num_eval_iters,
hooks=hooks,
eval_interval_secs=FLAGS.eval_interval_secs)
def main(unused_argv):
tf.logging.set_verbosity(tf.logging.INFO)
tf.set_random_seed(FLAGS.seed)
# Set up deployment (i.e., multi-GPUs and/or multi-replicas).
config = model_deploy.DeploymentConfig(num_clones=FLAGS.num_clones,
clone_on_cpu=FLAGS.clone_on_cpu,
replica_id=FLAGS.task,
num_replicas=FLAGS.num_replicas,
num_ps_tasks=FLAGS.num_ps_tasks)
# Split the batch across GPUs.
assert FLAGS.train_batch_size % config.num_clones == 0, (
'Training batch size not divisble by number of clones (GPUs).')
clone_batch_size = FLAGS.train_batch_size // config.num_clones
tf.gfile.MakeDirs(FLAGS.train_logdir)
tf.logging.info('Training segmentation and self-attention on %s set',
FLAGS.train_split)
if FLAGS.weakly:
tf.logging.info('Training classification on %s set',
FLAGS.train_split_cls)
else:
tf.logging.info('Training classification on %s set', FLAGS.train_split)
tf.logging.info('Enforcing consistency constraint on %s set',
FLAGS.train_split_cls)
with tf.Graph().as_default() as graph:
with tf.device(config.inputs_device()):
dataset = data_generator.Dataset(
dataset_name=FLAGS.dataset,
split_name=FLAGS.train_split,
dataset_dir=FLAGS.dataset_dir,
batch_size=clone_batch_size,
crop_size=[int(sz) for sz in FLAGS.train_crop_size],
min_resize_value=FLAGS.min_resize_value,
max_resize_value=FLAGS.max_resize_value,
resize_factor=FLAGS.resize_factor,
min_scale_factor=FLAGS.min_scale_factor,
max_scale_factor=FLAGS.max_scale_factor,
scale_factor_step_size=FLAGS.scale_factor_step_size,
model_variant=FLAGS.model_variant,
num_readers=4,
is_training=True,
should_shuffle=True,
should_repeat=True,
output_valid=True,
with_cls=True,
cls_only=False)
dataset_cls = data_generator.Dataset(
dataset_name=FLAGS.dataset,
split_name=FLAGS.train_split_cls,
dataset_dir=FLAGS.dataset_dir,
batch_size=clone_batch_size,
crop_size=[int(sz) for sz in FLAGS.train_crop_size],
min_resize_value=FLAGS.min_resize_value,
max_resize_value=FLAGS.max_resize_value,
resize_factor=FLAGS.resize_factor,
min_scale_factor=FLAGS.min_scale_factor,
max_scale_factor=FLAGS.max_scale_factor,
scale_factor_step_size=FLAGS.scale_factor_step_size,
model_variant=FLAGS.model_variant,
num_readers=4,
is_training=True,
should_shuffle=True,
should_repeat=True,
with_cls=FLAGS.weakly,
cls_only=False,
strong_weak=True)
# Create the global step on the device storing the variables.
with tf.device(config.variables_device()):
global_step = tf.train.get_or_create_global_step()
# Define the model and create clones.
model_fn = _build_pseudo_seg
model_args = (dataset.get_one_shot_iterator(),
dataset_cls.get_one_shot_iterator(), {
common.OUTPUT_TYPE: dataset.num_of_classes
}, dataset.ignore_label, clone_batch_size)
clones = model_deploy.create_clones(config,
model_fn,
args=model_args)
# Gather update_ops from the first clone. These contain, for example,
# the updates for the batch_norm variables created by model_fn.
first_clone_scope = config.clone_scope(0)
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS,
first_clone_scope)
# Gather initial summaries.
summaries = set(tf.get_collection(tf.GraphKeys.SUMMARIES))
# Add summaries for model variables.
for model_var in tf.model_variables():
summaries.add(tf.summary.histogram(model_var.op.name, model_var))
if FLAGS.use_attention:
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'max_prob_weak')).strip('/'))
summaries.add(tf.summary.histogram('max_prob_weak', summary))
summary = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, 'max_att_prob_weak')).strip('/'))
summaries.add(tf.summary.histogram('max_att_prob_weak', summary))
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'max_prob_avg')).strip('/'))
summaries.add(tf.summary.histogram('max_prob_avg', summary))
if FLAGS.soft_pseudo_label and FLAGS.temperature != 1.0:
summary = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, 'max_prob_avg_t')).strip('/'))
summaries.add(tf.summary.histogram('max_prob_avg_t', summary))
# Add summaries for images, labels, semantic predictions
# Visualize seg image and predictions
if FLAGS.save_summaries_images:
summary_image = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, common.IMAGE + '_seg')).strip('/'))
summaries.add(
tf.summary.image('samples/%s' % common.IMAGE + '_seg',
summary_image))
first_clone_label = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, common.LABEL + '_seg')).strip('/'))
# Scale up summary image pixel values for better visualization.
pixel_scaling = max(1, 255 // dataset.num_of_classes)
summary_label = tf.cast(first_clone_label * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('samples/%s' % common.LABEL + '_seg',
summary_label))
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, common.OUTPUT_TYPE + '_seg')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3), -1)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('samples/%s' % common.OUTPUT_TYPE + '_seg',
summary_predictions))
# For unlabeled image
summary_image = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'valid_mask')).strip('/'))
summaries.add(
tf.summary.image('sanity_check/valid_mask', summary_image))
summary_image = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'weak')).strip('/'))
summaries.add(tf.summary.image('unlabeled/weak', summary_image))
summary_image = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'strong')).strip('/'))
summaries.add(tf.summary.image('unlabeled/strong', summary_image))
first_clone_label = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'unlabeled')).strip('/'))
pixel_scaling = max(1, 255 // dataset.num_of_classes)
summary_label = tf.cast(first_clone_label * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('unlabeled/%s' % common.LABEL, summary_label))
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'logits_weak')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3), -1)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('unlabeled/logits_weak', summary_predictions))
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'logits_strong')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3), -1)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('unlabeled/logits_strong',
summary_predictions))
if FLAGS.use_attention:
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, 'att_logits_weak')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3),
-1)
predictions = tf.compat.v1.image.resize_bilinear(
predictions, [int(sz) for sz in FLAGS.train_crop_size],
align_corners=True)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('att/att_logits_weak',
summary_predictions))
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'cam_weak')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3),
-1)
predictions = tf.compat.v1.image.resize_bilinear(
predictions, [int(sz) for sz in FLAGS.train_crop_size],
align_corners=True)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('att/cam_weak', summary_predictions))
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, 'merged_logits')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3),
-1)
predictions = tf.compat.v1.image.resize_bilinear(
predictions, [int(sz) for sz in FLAGS.train_crop_size],
align_corners=True)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('att/merged_logits', summary_predictions))
first_clone_output = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, 'att_logits_labeled')).strip('/'))
predictions = tf.expand_dims(tf.argmax(first_clone_output, 3),
-1)
predictions = tf.compat.v1.image.resize_bilinear(
predictions, [int(sz) for sz in FLAGS.train_crop_size],
align_corners=True)
summary_predictions = tf.cast(predictions * pixel_scaling,
tf.uint8)
summaries.add(
tf.summary.image('att/att_logits_labeled',
summary_predictions))
# Add summaries for losses.
for loss in tf.get_collection(tf.GraphKeys.LOSSES, first_clone_scope):
summaries.add(tf.summary.scalar('losses/%s' % loss.op.name, loss))
# Monitor pseudo label quality
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'acc_seg')).strip('/'))
summaries.add(tf.summary.scalar('sanity_check/acc_seg', summary))
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'acc_weak')).strip('/'))
summaries.add(tf.summary.scalar('sanity_check/acc_weak', summary))
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'acc_strong')).strip('/'))
summaries.add(tf.summary.scalar('sanity_check/acc_strong', summary))
if FLAGS.pseudo_label_threshold > 0:
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'acc_pseudo')).strip('/'))
summaries.add(tf.summary.scalar('sanity_check/acc_pseudo',
summary))
summary = graph.get_tensor_by_name(
('%s/%s:0' %
(first_clone_scope, 'acc_strong_confident')).strip('/'))
summaries.add(
tf.summary.scalar('sanity_check/acc_strong_confident',
summary))
summary = graph.get_tensor_by_name(
('%s/%s:0' % (first_clone_scope, 'valid_ratio')).strip('/'))
summaries.add(
tf.summary.scalar('sanity_check/valid_ratio', summary))
# Build the optimizer based on the device specification.
with tf.device(config.optimizer_device()):
learning_rate = train_utils.get_model_learning_rate(
FLAGS.learning_policy,
FLAGS.base_learning_rate,
FLAGS.learning_rate_decay_step,
FLAGS.learning_rate_decay_factor,
FLAGS.training_number_of_steps,
FLAGS.learning_power,
FLAGS.slow_start_step,
FLAGS.slow_start_learning_rate,
decay_steps=FLAGS.decay_steps,
end_learning_rate=FLAGS.end_learning_rate)
summaries.add(tf.summary.scalar('learning_rate', learning_rate))
if FLAGS.optimizer == 'momentum':
optimizer = tf.train.MomentumOptimizer(learning_rate,
FLAGS.momentum)
elif FLAGS.optimizer == 'adam':
optimizer = tf.train.AdamOptimizer(
learning_rate=FLAGS.adam_learning_rate,
epsilon=FLAGS.adam_epsilon)
else:
raise ValueError('Unknown optimizer')
startup_delay_steps = FLAGS.task * FLAGS.startup_delay_steps
with tf.device(config.variables_device()):
total_loss, grads_and_vars = model_deploy.optimize_clones(
clones, optimizer)
total_loss = tf.check_numerics(total_loss, 'Loss is inf or nan.')
summaries.add(tf.summary.scalar('total_loss', total_loss))
# Modify the gradients for biases and last layer variables.
last_layers = model.get_extra_layer_scopes(
FLAGS.last_layers_contain_logits_only)
grad_mult = train_utils.get_model_gradient_multipliers(
last_layers, FLAGS.last_layer_gradient_multiplier)
if grad_mult:
grads_and_vars = slim.learning.multiply_gradients(
grads_and_vars, grad_mult)
# NOTE: Neither last cls nor last seg layer loads pre-trained weights
last_layers += [
'{}/logits'.format(FLAGS.model_variant).replace('_beta', '')
]
# Create gradient update op.
grad_updates = optimizer.apply_gradients(grads_and_vars,
global_step=global_step)
update_ops.append(grad_updates)
update_op = tf.group(*update_ops)
with tf.control_dependencies([update_op]):
train_tensor = tf.identity(total_loss, name='train_op')
# Add the summaries from the first clone. These contain the summaries
# created by model_fn and either optimize_clones() or _gather_clone_loss().
summaries |= set(
tf.get_collection(tf.GraphKeys.SUMMARIES, first_clone_scope))
# Merge all summaries together.
summary_op = tf.summary.merge(list(summaries))
# Soft placement allows placing on CPU ops without GPU implementation.
session_config = tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False)
session_config.gpu_options.allow_growth = True
# Start the training.
profile_dir = FLAGS.profile_logdir
if profile_dir is not None:
tf.gfile.MakeDirs(profile_dir)
with contrib_tfprof.ProfileContext(enabled=profile_dir is not None,
profile_dir=profile_dir):
init_fn = None
if FLAGS.tf_initial_checkpoint:
init_fn = train_utils.get_model_init_fn(
FLAGS.train_logdir,
FLAGS.tf_initial_checkpoint,
FLAGS.initialize_last_layer,
last_layers,
ignore_missing_vars=True)
slim.learning.train(train_tensor,
logdir=FLAGS.train_logdir,
log_every_n_steps=FLAGS.log_steps,
master=FLAGS.master,
number_of_steps=FLAGS.training_number_of_steps,
is_chief=(FLAGS.task == 0),
session_config=session_config,
startup_delay_steps=startup_delay_steps,
init_fn=init_fn,
summary_op=summary_op,
save_summaries_secs=FLAGS.save_summaries_secs,
save_interval_secs=FLAGS.save_interval_secs)
def main(unused_argv):
tf.logging.set_verbosity(tf.logging.INFO)
# Get dataset-dependent information.
dataset = data_generator.Dataset(
dataset_name=FLAGS.dataset,
split_name=FLAGS.vis_split,
dataset_dir=FLAGS.dataset_dir,
batch_size=FLAGS.vis_batch_size,
crop_size=[int(sz) for sz in FLAGS.vis_crop_size],
min_resize_value=FLAGS.min_resize_value,
max_resize_value=FLAGS.max_resize_value,
resize_factor=FLAGS.resize_factor,
model_variant=FLAGS.model_variant,
is_training=False,
should_shuffle=False,
should_repeat=False)
train_id_to_eval_id = None
if dataset.dataset_name == data_generator.get_cityscapes_dataset_name():
tf.logging.info('Cityscapes requires converting train_id to eval_id.')
train_id_to_eval_id = _CITYSCAPES_TRAIN_ID_TO_EVAL_ID
# Prepare for visualization.
tf.gfile.MakeDirs(FLAGS.vis_logdir)
save_dir = os.path.join(FLAGS.vis_logdir, _SEMANTIC_PREDICTION_SAVE_FOLDER)
tf.gfile.MakeDirs(save_dir)
raw_save_dir = os.path.join(FLAGS.vis_logdir,
_RAW_SEMANTIC_PREDICTION_SAVE_FOLDER)
tf.gfile.MakeDirs(raw_save_dir)
tf.logging.info('Visualizing on %s set', FLAGS.vis_split)
with tf.Graph().as_default():
samples = dataset.get_one_shot_iterator().get_next()
model_options = common.ModelOptions(
outputs_to_num_classes={
common.OUTPUT_TYPE: dataset.num_of_classes
},
crop_size=[int(sz) for sz in FLAGS.vis_crop_size],
atrous_rates=FLAGS.atrous_rates,
output_stride=FLAGS.output_stride)
if tuple(FLAGS.eval_scales) == (1.0, ):
tf.logging.info('Performing single-scale test.')
predictions = model.predict_labels(
samples[common.IMAGE],
model_options=model_options,
image_pyramid=FLAGS.image_pyramid)
else:
tf.logging.info('Performing multi-scale test.')
if FLAGS.quantize_delay_step >= 0:
raise ValueError(
'Quantize mode is not supported with multi-scale test.')
predictions = model.predict_labels_multi_scale(
samples[common.IMAGE],
model_options=model_options,
eval_scales=FLAGS.eval_scales,
add_flipped_images=FLAGS.add_flipped_images)
predictions = predictions[common.OUTPUT_TYPE]
if FLAGS.min_resize_value and FLAGS.max_resize_value:
# Only support batch_size = 1, since we assume the dimensions of original
# image after tf.squeeze is [height, width, 3].
assert FLAGS.vis_batch_size == 1
# Reverse the resizing and padding operations performed in preprocessing.
# First, we slice the valid regions (i.e., remove padded region) and then
# we resize the predictions back.
original_image = tf.squeeze(samples[common.ORIGINAL_IMAGE])
original_image_shape = tf.shape(original_image)
predictions = tf.slice(
predictions, [0, 0, 0],
[1, original_image_shape[0], original_image_shape[1]])
resized_shape = tf.to_int32([
tf.squeeze(samples[common.HEIGHT]),
tf.squeeze(samples[common.WIDTH])
])
predictions = tf.squeeze(
tf.image.resize_images(
tf.expand_dims(predictions, 3),
resized_shape,
method=tf.image.ResizeMethod.NEAREST_NEIGHBOR,
align_corners=True), 3)
tf.train.get_or_create_global_step()
if FLAGS.quantize_delay_step >= 0:
contrib_quantize.create_eval_graph()
num_iteration = 0
max_num_iteration = FLAGS.max_number_of_iterations
if True:
checkpoint_path = FLAGS.checkpoint_dir
num_iteration += 1
tf.logging.info('Starting visualization at ' +
time.strftime('%Y-%m-%d-%H:%M:%S', time.gmtime()))
tf.logging.info('Visualizing with model %s', checkpoint_path)
scaffold = tf.train.Scaffold(
init_op=tf.global_variables_initializer())
session_creator = tf.train.ChiefSessionCreator(
scaffold=scaffold,
master=FLAGS.master,
checkpoint_filename_with_path=checkpoint_path)
with tf.train.MonitoredSession(session_creator=session_creator,
hooks=None) as sess:
batch = 0
image_id_offset = 0
while not sess.should_stop():
tf.logging.info('Visualizing batch %d', batch + 1)
_process_batch(
sess=sess,
original_images=samples[common.ORIGINAL_IMAGE],
semantic_predictions=predictions,
image_names=samples[common.IMAGE_NAME],
image_heights=samples[common.HEIGHT],
image_widths=samples[common.WIDTH],
image_id_offset=image_id_offset,
save_dir=save_dir,
raw_save_dir=raw_save_dir,
train_id_to_eval_id=train_id_to_eval_id)
image_id_offset += FLAGS.vis_batch_size
batch += 1
tf.logging.info('Finished visualization at ' +
time.strftime('%Y-%m-%d-%H:%M:%S', time.gmtime()))