Merge branch 'classification_application' into 'dev'

better tensorboard metrics, pylint fixes

See merge request CMIC/NiftyNet!194

niftynet/application/classification_application.py

@@ -1,3 +1,10 @@
+"""
+This module defines an image-level classification application
+that maps from images to scalar, multi-class labels.
+
+This class is instantiated and initalized by the application_driver.
+"""
+
 import tensorflow as tf
 
 from niftynet.application.base_application import BaseApplication
@@ -6,7 +13,8 @@
 from niftynet.engine.application_variables import \
     CONSOLE, NETWORK_OUTPUT, TF_SUMMARIES
 from niftynet.engine.sampler_resize import ResizeSampler
-from niftynet.engine.windows_aggregator_classifier import ClassifierSamplesAggregator
+from niftynet.engine.windows_aggregator_classifier import \
+    ClassifierSamplesAggregator
 from niftynet.io.image_reader import ImageReader
 from niftynet.layer.discrete_label_normalisation import \
     DiscreteLabelNormalisationLayer
@@ -25,6 +33,16 @@
 
 
 class ClassificationApplication(BaseApplication):
+    """This class defines an application for image-level classification
+    problems mapping from images to scalar labels.
+
+    This is the application class to be instantiated by the driver
+    and referred to in configuration files.
+
+    Although structurally similar to segmentation, this application
+    supports different samplers/aggregators (because patch-based
+    processing is not appropriate), and monitoring metrics."""
+
     REQUIRED_CONFIG_SECTION = "CLASSIFICATION"
 
     def __init__(self, net_param, action_param, is_training):
@@ -140,7 +158,7 @@ def initialise_resize_sampler(self):
             batch_size=self.net_param.batch_size,
             shuffle_buffer=self.is_training,
             queue_length=self.net_param.queue_length) for reader in
-            self.readers]]
+                         self.readers]]
 
     def initialise_aggregator(self):
         self.output_decoder = ClassifierSamplesAggregator(
@@ -177,6 +195,49 @@ def initialise_network(self):
             b_regularizer=b_regularizer,
             acti_func=self.net_param.activation_function)
 
+    def add_confusion_matrix_summaries_(self,
+                                        outputs_collector,
+                                        net_out,
+                                        data_dict):
+        """ This method defines several monitoring metrics that
+        are derived from the confusion matrix """
+        labels = tf.reshape(tf.cast(data_dict['label'], tf.int64), [-1])
+        prediction = tf.reshape(tf.argmax(net_out, -1), [-1])
+        num_classes = self.classification_param.num_classes
+        conf_mat = tf.contrib.metrics.confusion_matrix(labels,
+                                                       prediction,
+                                                       num_classes)
+        conf_mat = tf.to_float(conf_mat) / float(self.net_param.batch_size)
+        if self.classification_param.num_classes == 2:
+            outputs_collector.add_to_collection(
+                var=conf_mat[1][1], name='true_positives',
+                average_over_devices=True, summary_type='scalar',
+                collection=TF_SUMMARIES)
+            outputs_collector.add_to_collection(
+                var=conf_mat[1][0], name='false_negatives',
+                average_over_devices=True, summary_type='scalar',
+                collection=TF_SUMMARIES)
+            outputs_collector.add_to_collection(
+                var=conf_mat[0][1], name='false_positives',
+                average_over_devices=True, summary_type='scalar',
+                collection=TF_SUMMARIES)
+            outputs_collector.add_to_collection(
+                var=conf_mat[0][0], name='true_negatives',
+                average_over_devices=True, summary_type='scalar',
+                collection=TF_SUMMARIES)
+        else:
+            outputs_collector.add_to_collection(
+                var=conf_mat[tf.newaxis, :, :, tf.newaxis],
+                name='confusion_matrix',
+                average_over_devices=True, summary_type='image',
+                collection=TF_SUMMARIES)
+
+        outputs_collector.add_to_collection(
+            var=tf.trace(conf_mat), name='accuracy',
+            average_over_devices=True, summary_type='scalar',
+            collection=TF_SUMMARIES)
+
+
     def connect_data_and_network(self,
                                  outputs_collector=None,
                                  gradients_collector=None):
@@ -225,6 +286,9 @@ def switch_sampler(for_training):
                 var=data_loss, name='data_loss',
                 average_over_devices=True, summary_type='scalar',
                 collection=TF_SUMMARIES)
+            self.add_confusion_matrix_summaries_(outputs_collector,
+                                                 net_out,
+                                                 data_dict)
         else:
             # converting logits into final output for
             # classification probabilities or argmax classification labels