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/datashuffler/TFRecord.py

@@ -16,15 +16,19 @@ class TFRecord(object):
     Datashuffler that wraps tfrecords    
     """
 
-
     def __init__(self,filename_queue,
                          input_shape=[None, 28, 28, 1],
-                         input_dtype="float32",
+                         output_shape=[None, 28, 28, 1],
+                         input_dtype=tf.float32,
                          batch_size=32,
                          seed=10,
                          prefetch_capacity=50,
                          prefetch_threads=5, 
-                         shuffle=True):
+                         shuffle=True,
+                         normalization=False,
+                         random_flip=True,
+                         random_crop=True
+                         ):
 
         # Setting the seed for the pseudo random number generator
         self.seed = seed
@@ -38,6 +42,7 @@ class TFRecord(object):
         # Preparing the inputs
         self.filename_queue = filename_queue
         self.input_shape = tuple(input_shape)
+        self.output_shape = output_shape
 
         # Prefetch variables
         self.prefetch = True
@@ -48,6 +53,8 @@ class TFRecord(object):
         self.label_ph = None
         
         self.shuffle = shuffle
+        self.normalization = normalization
+        self.random_crop = random_crop
 
     def __call__(self, element, from_queue=False):
         """
@@ -83,7 +90,7 @@ class TFRecord(object):
         features = tf.parse_single_example(serialized_example, features=feature)
         
         # Convert the image data from string back to the numbers
-        image = tf.decode_raw(features['train/data'], tf.float32)
+        image = tf.decode_raw(features['train/data'], self.input_dtype)
         #image = tf.decode_raw(features['train/data'], tf.uint8)
         
         # Cast label data into int32
@@ -91,7 +98,19 @@ class TFRecord(object):
         
         # Reshape image data into the original shape
         image = tf.reshape(image, self.input_shape[1:])
+
+        # Casting to float32
+        image = tf.cast(image, tf.float32)
         
+        if self.random_crop:
+            image = tf.image.resize_image_with_crop_or_pad(image, self.output_shape[1], self.output_shape[2])
+
+        # normalizing data
+        if self.normalization:
+            image = tf.image.per_image_standardization(image)
+
+        image.set_shape(tuple(self.output_shape[1:]))
+
         if self.shuffle:
             data_ph, label_ph = tf.train.shuffle_batch([image, label], batch_size=self.batch_size,
                              capacity=self.prefetch_capacity, num_threads=self.prefetch_threads,
@@ -101,7 +120,6 @@ class TFRecord(object):
                              capacity=self.prefetch_capacity, num_threads=self.prefetch_threads, name="batch")
         
         
-        
         self.data_ph = data_ph
         self.label_ph = label_ph