Updates

• config file from train.py accepts tensors as input
• TFRecord, add keyword argument to flag the shuffling mechanism
• Added InceptionResnetV1 in our basket of networks
• Added an option to sum the regularization losses in the CrossEntropy loss
• Implemented validation with embeddings (officially appending to the tensorboard)
• Created method Trainer.load_variables_from_external_model that load variables from another model in the current graph
• Implemented MeanSoftMaxLossCenterLoss (http://ydwen.github.io/papers/WenECCV16.pdf)
• Patched InceptionV2 (Added the flag trainable for each variable in the model)
• Implemented TFRecordImage (read, data augmentation, scaling and casting to float32)
• Added data type as input in the TFRecord datashufflers

bob/learn/tensorflow/loss/BaseLoss.py

@@ -31,7 +31,7 @@ class MeanSoftMaxLoss(object):
     Mean softmax loss. Basically it wrapps the function tf.nn.sparse_softmax_cross_entropy_with_logits.
     """
 
-    def __init__(self, name="loss"):
+    def __init__(self, name="loss", add_regularization_losses=True):
         """
         Constructor
         
@@ -43,8 +43,15 @@ class MeanSoftMaxLoss(object):
         """
     
         self.name = name
+        self.add_regularization_losses = add_regularization_losses
 
     def __call__(self, graph, label):
     
-        return tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(
-                              logits=graph, labels=label), name=self.name)
+        loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(
+                                          logits=graph, labels=label), name=self.name)
+    
+        if self.add_regularization_losses:
+            regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)
+            return tf.add_n([loss] + regularization_losses, name='total_loss')
+        else:
+            return loss