Multilabel classifications scripts: adding TODO

Author: aelbouchti

estimator_multiclass.py

@@ -89,12 +89,16 @@ def model_fn(features, mode):
         tf.train.init_from_checkpoint(checkpoint_file, 
                             {v.name.split(':')[0]: v for v in variables_to_restore})
 
-        #Defining losses and regulization ops:
+    #Defining losses and regulization ops:
     with tf.name_scope("loss_op"):
         loss = tf.losses.sigmoid_cross_entropy(multi_class_labels = features['image/class/id'], logits = logits)
         total_loss = tf.losses.get_total_loss() #obtain the regularization losses as well
     #FIXME: Replace classifier function (sigmoid / softmax)
-    print(features['image/class/id'])
+   
+    #TODO: Add a func to transform logit tensor to a label-like tensor
+    # If value[][class_id]<0.5 then value[][class_id] = 0. else value[][class_id]= 1.
+    #It is necessary for a multilabel classification problem
+    
     if mode != tf.estimator.ModeKeys.PREDICT:
         metrics = {
         'Accuracy': tf.metrics.accuracy(features['image/class/id'], logits, name="acc_op"),