# -*- coding: utf-8 -*-
import tensorflow as tf
from niftynet.application.base_application import BaseApplication
from niftynet.engine.application_factory import \
ApplicationNetFactory, InitializerFactory, OptimiserFactory
from niftynet.engine.application_variables import \
CONSOLE, NETWORK_OUTPUT, TF_SUMMARIES
from niftynet.engine.sampler_grid_v2 import GridSampler
from niftynet.engine.sampler_resize_v2 import ResizeSampler
from niftynet.engine.sampler_uniform_v2 import UniformSampler
from niftynet.engine.sampler_weighted_v2 import WeightedSampler
from niftynet.engine.sampler_balanced_v2 import BalancedSampler
from niftynet.engine.windows_aggregator_grid import GridSamplesAggregator
from niftynet.engine.windows_aggregator_resize import ResizeSamplesAggregator
from niftynet.io.image_reader import ImageReader
from niftynet.layer.binary_masking import BinaryMaskingLayer
from niftynet.layer.discrete_label_normalisation import \
DiscreteLabelNormalisationLayer
from niftynet.layer.histogram_normalisation import \
HistogramNormalisationLayer
from niftynet.layer.loss_segmentation import LossFunction
from niftynet.layer.mean_variance_normalisation import \
MeanVarNormalisationLayer
from niftynet.layer.pad import PadLayer
from niftynet.layer.post_processing import PostProcessingLayer
from niftynet.layer.rand_flip import RandomFlipLayer
from niftynet.layer.rand_rotation import RandomRotationLayer
from niftynet.layer.rand_spatial_scaling import RandomSpatialScalingLayer
from niftynet.layer.rgb_histogram_equilisation import \
RGBHistogramEquilisationLayer
from niftynet.evaluation.segmentation_evaluator import SegmentationEvaluator
from niftynet.layer.rand_elastic_deform import RandomElasticDeformationLayer
SUPPORTED_INPUT = set(
['image', 'label', 'weight', 'sampler', 'inferred', 'value'])
[docs]class SegmentationApplication(BaseApplication):
REQUIRED_CONFIG_SECTION = "SEGMENTATION"
def __init__(self, net_param, action_param, action):
super(SegmentationApplication, self).__init__()
tf.logging.info('starting segmentation application')
self.action = action
self.net_param = net_param
self.action_param = action_param
self.data_param = None
self.segmentation_param = None
self.SUPPORTED_SAMPLING = {
'uniform': (self.initialise_uniform_sampler,
self.initialise_grid_sampler,
self.initialise_grid_aggregator),
'weighted': (self.initialise_weighted_sampler,
self.initialise_grid_sampler,
self.initialise_grid_aggregator),
'resize': (self.initialise_resize_sampler,
self.initialise_resize_sampler,
self.initialise_resize_aggregator),
'balanced': (self.initialise_balanced_sampler,
self.initialise_grid_sampler,
self.initialise_grid_aggregator),
}
[docs] def initialise_dataset_loader(
self, data_param=None, task_param=None, data_partitioner=None):
self.data_param = data_param
self.segmentation_param = task_param
# initialise input image readers
if self.is_training:
reader_names = ('image', 'label', 'weight', 'sampler')
elif self.is_inference:
# in the inference process use `image` input only
reader_names = ('image',)
elif self.is_evaluation:
reader_names = ('image', 'label', 'inferred')
else:
tf.logging.fatal(
'Action `%s` not supported. Expected one of %s',
self.action, self.SUPPORTED_PHASES)
raise ValueError
try:
reader_phase = self.action_param.dataset_to_infer
except AttributeError:
reader_phase = None
file_lists = data_partitioner.get_file_lists_by(
phase=reader_phase, action=self.action)
self.readers = [
ImageReader(reader_names).initialise(
data_param, task_param, file_list) for file_list in file_lists]
# initialise input preprocessing layers
foreground_masking_layer = BinaryMaskingLayer(
type_str=self.net_param.foreground_type,
multimod_fusion=self.net_param.multimod_foreground_type,
threshold=0.0) \
if self.net_param.normalise_foreground_only else None
mean_var_normaliser = MeanVarNormalisationLayer(
image_name='image', binary_masking_func=foreground_masking_layer) \
if self.net_param.whitening else None
histogram_normaliser = HistogramNormalisationLayer(
image_name='image',
modalities=vars(task_param).get('image'),
model_filename=self.net_param.histogram_ref_file,
binary_masking_func=foreground_masking_layer,
norm_type=self.net_param.norm_type,
cutoff=self.net_param.cutoff,
name='hist_norm_layer') \
if (self.net_param.histogram_ref_file and
self.net_param.normalisation) else None
rgb_normaliser = RGBHistogramEquilisationLayer(
image_name='image',
name='rbg_norm_layer') if self.net_param.rgb_normalisation else None
label_normalisers = None
if self.net_param.histogram_ref_file and \
task_param.label_normalisation:
label_normalisers = [DiscreteLabelNormalisationLayer(
image_name='label',
modalities=vars(task_param).get('label'),
model_filename=self.net_param.histogram_ref_file)]
if self.is_evaluation:
label_normalisers.append(
DiscreteLabelNormalisationLayer(
image_name='inferred',
modalities=vars(task_param).get('inferred'),
model_filename=self.net_param.histogram_ref_file))
label_normalisers[-1].key = label_normalisers[0].key
normalisation_layers = []
if histogram_normaliser is not None:
normalisation_layers.append(histogram_normaliser)
if rgb_normaliser is not None:
normalisation_layers.append(rgb_normaliser)
if mean_var_normaliser is not None:
normalisation_layers.append(mean_var_normaliser)
if task_param.label_normalisation and \
(self.is_training or not task_param.output_prob):
normalisation_layers.extend(label_normalisers)
volume_padding_layer = [PadLayer(
image_name=SUPPORTED_INPUT,
border=self.net_param.volume_padding_size,
mode=self.net_param.volume_padding_mode,
pad_to=self.net_param.volume_padding_to_size)
]
# initialise training data augmentation layers
augmentation_layers = []
if self.is_training:
train_param = self.action_param
self.patience = train_param.patience
self.mode = self.action_param.early_stopping_mode
if train_param.random_flipping_axes != -1:
augmentation_layers.append(RandomFlipLayer(
flip_axes=train_param.random_flipping_axes))
if train_param.scaling_percentage:
augmentation_layers.append(RandomSpatialScalingLayer(
min_percentage=train_param.scaling_percentage[0],
max_percentage=train_param.scaling_percentage[1],
antialiasing=train_param.antialiasing,
isotropic=train_param.isotropic_scaling))
if train_param.rotation_angle or \
train_param.rotation_angle_x or \
train_param.rotation_angle_y or \
train_param.rotation_angle_z:
rotation_layer = RandomRotationLayer()
if train_param.rotation_angle:
rotation_layer.init_uniform_angle(
train_param.rotation_angle)
else:
rotation_layer.init_non_uniform_angle(
train_param.rotation_angle_x,
train_param.rotation_angle_y,
train_param.rotation_angle_z)
augmentation_layers.append(rotation_layer)
if train_param.do_elastic_deformation:
spatial_rank = list(self.readers[0].spatial_ranks.values())[0]
augmentation_layers.append(RandomElasticDeformationLayer(
spatial_rank=spatial_rank,
num_controlpoints=train_param.num_ctrl_points,
std_deformation_sigma=train_param.deformation_sigma,
proportion_to_augment=train_param.proportion_to_deform))
# only add augmentation to first reader (not validation reader)
self.readers[0].add_preprocessing_layers(
volume_padding_layer + normalisation_layers + augmentation_layers)
for reader in self.readers[1:]:
reader.add_preprocessing_layers(
volume_padding_layer + normalisation_layers)
# Checking num_classes is set correctly
if self.segmentation_param.num_classes <= 1:
raise ValueError("Number of classes must be at least 2 for segmentation")
for preprocessor in self.readers[0].preprocessors:
if preprocessor.name == 'label_norm':
if len(preprocessor.label_map[preprocessor.key[0]]) != self.segmentation_param.num_classes:
raise ValueError("Number of unique labels must be equal to "
"number of classes (check histogram_ref file)")
[docs] def initialise_weighted_sampler(self):
self.sampler = [[WeightedSampler(
reader=reader,
window_sizes=self.data_param,
batch_size=self.net_param.batch_size,
windows_per_image=self.action_param.sample_per_volume,
queue_length=self.net_param.queue_length) for reader in
self.readers]]
[docs] def initialise_resize_sampler(self):
self.sampler = [[ResizeSampler(
reader=reader,
window_sizes=self.data_param,
batch_size=self.net_param.batch_size,
shuffle=self.is_training,
smaller_final_batch_mode=self.net_param.smaller_final_batch_mode,
queue_length=self.net_param.queue_length) for reader in
self.readers]]
[docs] def initialise_grid_sampler(self):
self.sampler = [[GridSampler(
reader=reader,
window_sizes=self.data_param,
batch_size=self.net_param.batch_size,
spatial_window_size=self.action_param.spatial_window_size,
window_border=self.action_param.border,
smaller_final_batch_mode=self.net_param.smaller_final_batch_mode,
queue_length=self.net_param.queue_length) for reader in
self.readers]]
[docs] def initialise_balanced_sampler(self):
self.sampler = [[BalancedSampler(
reader=reader,
window_sizes=self.data_param,
batch_size=self.net_param.batch_size,
windows_per_image=self.action_param.sample_per_volume,
queue_length=self.net_param.queue_length) for reader in
self.readers]]
[docs] def initialise_grid_aggregator(self):
self.output_decoder = GridSamplesAggregator(
image_reader=self.readers[0],
output_path=self.action_param.save_seg_dir,
window_border=self.action_param.border,
interp_order=self.action_param.output_interp_order,
postfix=self.action_param.output_postfix,
fill_constant=self.action_param.fill_constant)
[docs] def initialise_resize_aggregator(self):
self.output_decoder = ResizeSamplesAggregator(
image_reader=self.readers[0],
output_path=self.action_param.save_seg_dir,
window_border=self.action_param.border,
interp_order=self.action_param.output_interp_order,
postfix=self.action_param.output_postfix)
[docs] def initialise_sampler(self):
if self.is_training:
self.SUPPORTED_SAMPLING[self.net_param.window_sampling][0]()
elif self.is_inference:
self.SUPPORTED_SAMPLING[self.net_param.window_sampling][1]()
[docs] def initialise_aggregator(self):
self.SUPPORTED_SAMPLING[self.net_param.window_sampling][2]()
[docs] def initialise_network(self):
w_regularizer = None
b_regularizer = None
reg_type = self.net_param.reg_type.lower()
decay = self.net_param.decay
if reg_type == 'l2' and decay > 0:
from tensorflow.contrib.layers.python.layers import regularizers
w_regularizer = regularizers.l2_regularizer(decay)
b_regularizer = regularizers.l2_regularizer(decay)
elif reg_type == 'l1' and decay > 0:
from tensorflow.contrib.layers.python.layers import regularizers
w_regularizer = regularizers.l1_regularizer(decay)
b_regularizer = regularizers.l1_regularizer(decay)
self.net = ApplicationNetFactory.create(self.net_param.name)(
num_classes=self.segmentation_param.num_classes,
w_initializer=InitializerFactory.get_initializer(
name=self.net_param.weight_initializer),
b_initializer=InitializerFactory.get_initializer(
name=self.net_param.bias_initializer),
w_regularizer=w_regularizer,
b_regularizer=b_regularizer,
acti_func=self.net_param.activation_function)
[docs] def connect_data_and_network(self,
outputs_collector=None,
gradients_collector=None):
def switch_sampler(for_training):
with tf.name_scope('train' if for_training else 'validation'):
sampler = self.get_sampler()[0][0 if for_training else -1]
return sampler.pop_batch_op()
def mixup_switch_sampler(for_training):
# get first set of samples
d_dict = switch_sampler(for_training=for_training)
mix_fields = ('image', 'weight', 'label')
if not for_training:
with tf.name_scope('nomix'):
# ensure label is appropriate for dense loss functions
ground_truth = tf.cast(d_dict['label'], tf.int32)
one_hot = tf.one_hot(tf.squeeze(ground_truth, axis=-1),
depth=self.segmentation_param.num_classes)
d_dict['label'] = one_hot
else:
with tf.name_scope('mixup'):
# get the mixing parameter from the Beta distribution
alpha = self.segmentation_param.mixup_alpha
beta = tf.distributions.Beta(alpha, alpha) # 1, 1: uniform:
rand_frac = beta.sample()
# get another minibatch
d_dict_to_mix = switch_sampler(for_training=True)
# look at binarised labels: sort them
if self.segmentation_param.mix_match:
# sum up the positive labels to sort by their volumes
inds1 = tf.argsort(tf.map_fn(tf.reduce_sum, tf.cast(d_dict['label'], tf.int64)))
inds2 = tf.argsort(tf.map_fn(tf.reduce_sum, tf.cast(d_dict_to_mix['label'] > 0, tf.int64)))
for field in [field for field in mix_fields if field in d_dict]:
d_dict[field] = tf.gather(d_dict[field], indices=inds1)
# note: sorted for opposite directions for d_dict_to_mix
d_dict_to_mix[field] = tf.gather(d_dict_to_mix[field], indices=inds2[::-1])
# making the labels dense and one-hot
for d in (d_dict, d_dict_to_mix):
ground_truth = tf.cast(d['label'], tf.int32)
one_hot = tf.one_hot(tf.squeeze(ground_truth, axis=-1),
depth=self.segmentation_param.num_classes)
d['label'] = one_hot
# do the mixing for any fields that are relevant and present
mixed_up = {field: d_dict[field] * rand_frac + d_dict_to_mix[field] * (1 - rand_frac) for field
in mix_fields if field in d_dict}
# reassign all relevant values in d_dict
d_dict.update(mixed_up)
return d_dict
if self.is_training:
if not self.segmentation_param.do_mixup:
data_dict = tf.cond(tf.logical_not(self.is_validation),
lambda: switch_sampler(for_training=True),
lambda: switch_sampler(for_training=False))
else:
# mix up the samples if not in validation phase
data_dict = tf.cond(tf.logical_not(self.is_validation),
lambda: mixup_switch_sampler(for_training=True),
lambda: mixup_switch_sampler(for_training=False)) # don't mix the validation
image = tf.cast(data_dict['image'], tf.float32)
net_args = {'is_training': self.is_training,
'keep_prob': self.net_param.keep_prob}
net_out = self.net(image, **net_args)
with tf.name_scope('Optimiser'):
optimiser_class = OptimiserFactory.create(
name=self.action_param.optimiser)
self.optimiser = optimiser_class.get_instance(
learning_rate=self.action_param.lr)
loss_func = LossFunction(
n_class=self.segmentation_param.num_classes,
loss_type=self.action_param.loss_type,
softmax=self.segmentation_param.softmax)
data_loss = loss_func(
prediction=net_out,
ground_truth=data_dict.get('label', None),
weight_map=data_dict.get('weight', None))
reg_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)
if self.net_param.decay > 0.0 and reg_losses:
reg_loss = tf.reduce_mean(
[tf.reduce_mean(reg_loss) for reg_loss in reg_losses])
loss = data_loss + reg_loss
else:
loss = data_loss
# Get all vars
to_optimise = tf.trainable_variables()
vars_to_freeze = \
self.action_param.vars_to_freeze or \
self.action_param.vars_to_restore
if vars_to_freeze:
import re
var_regex = re.compile(vars_to_freeze)
# Only optimise vars that are not frozen
to_optimise = \
[v for v in to_optimise if not var_regex.search(v.name)]
tf.logging.info(
"Optimizing %d out of %d trainable variables, "
"the other variables fixed (--vars_to_freeze %s)",
len(to_optimise),
len(tf.trainable_variables()),
vars_to_freeze)
grads = self.optimiser.compute_gradients(
loss, var_list=to_optimise, colocate_gradients_with_ops=True)
self.total_loss = loss
# collecting gradients variables
gradients_collector.add_to_collection([grads])
# collecting output variables
outputs_collector.add_to_collection(
var=self.total_loss, name='total_loss',
average_over_devices=True, collection=CONSOLE)
outputs_collector.add_to_collection(
var=self.total_loss, name='total_loss',
average_over_devices=True, summary_type='scalar',
collection=TF_SUMMARIES)
outputs_collector.add_to_collection(
var=data_loss, name='loss',
average_over_devices=False, collection=CONSOLE)
outputs_collector.add_to_collection(
var=data_loss, name='loss',
average_over_devices=True, summary_type='scalar',
collection=TF_SUMMARIES)
# outputs_collector.add_to_collection(
# var=image*180.0, name='image',
# average_over_devices=False, summary_type='image3_sagittal',
# collection=TF_SUMMARIES)
# outputs_collector.add_to_collection(
# var=image, name='image',
# average_over_devices=False,
# collection=NETWORK_OUTPUT)
# outputs_collector.add_to_collection(
# var=tf.reduce_mean(image), name='mean_image',
# average_over_devices=False, summary_type='scalar',
# collection=CONSOLE)
elif self.is_inference:
# converting logits into final output for
# classification probabilities or argmax classification labels
data_dict = switch_sampler(for_training=False)
image = tf.cast(data_dict['image'], tf.float32)
net_args = {'is_training': self.is_training,
'keep_prob': self.net_param.keep_prob}
net_out = self.net(image, **net_args)
output_prob = self.segmentation_param.output_prob
num_classes = self.segmentation_param.num_classes
if output_prob and num_classes > 1:
post_process_layer = PostProcessingLayer(
'SOFTMAX', num_classes=num_classes)
elif not output_prob and num_classes > 1:
post_process_layer = PostProcessingLayer(
'ARGMAX', num_classes=num_classes)
else:
post_process_layer = PostProcessingLayer(
'IDENTITY', num_classes=num_classes)
net_out = post_process_layer(net_out)
outputs_collector.add_to_collection(
var=net_out, name='window',
average_over_devices=False, collection=NETWORK_OUTPUT)
outputs_collector.add_to_collection(
var=data_dict['image_location'], name='location',
average_over_devices=False, collection=NETWORK_OUTPUT)
self.initialise_aggregator()
[docs] def interpret_output(self, batch_output):
if self.is_inference:
return self.output_decoder.decode_batch(
{'window_seg': batch_output['window']},
batch_output['location'])
return True
[docs] def initialise_evaluator(self, eval_param):
self.eval_param = eval_param
self.evaluator = SegmentationEvaluator(self.readers[0],
self.segmentation_param,
eval_param)
[docs] def add_inferred_output(self, data_param, task_param):
return self.add_inferred_output_like(data_param, task_param, 'label')
