batch_norm and xavier init added

core/model.py

@@ -48,14 +48,14 @@ def __init__(self, word_to_idx, dim_feature=[196, 512], dim_embed=512, dim_hidde
         self._start = word_to_idx['<START>']
         self._null = word_to_idx['<NULL>']
 
-        self.weight_initializer = tf.random_normal_initializer(mean=0.0, stddev=0.01)
+        self.weight_initializer = tf.contrib.layers.xavier_initializer()
         self.const_initializer = tf.constant_initializer(0.0)
         self.emb_initializer = tf.random_uniform_initializer(minval=-1.0, maxval=1.0)
 
         # Place holder for features and captions
         self.features = tf.placeholder(tf.float32, [None, self.L, self.D])
         self.captions = tf.placeholder(tf.int32, [None, self.T + 1])
-
+    
     def _get_initial_lstm(self, features):
         with tf.variable_scope('initial_lstm'):
             features_mean = tf.reduce_mean(features, 1)
@@ -126,6 +126,15 @@ def _decode_lstm(self, x, h, context, dropout=False, reuse=False):
                 h_logits = tf.nn.dropout(h_logits, 0.5)
             out_logits = tf.matmul(h_logits, w_out) + b_out
             return out_logits
+
+    def _batch_norm(self, x, mode='train', name=None):
+        return tf.contrib.layers.batch_norm(inputs=x, 
+                                            decay=0.95,
+                                            center=True,
+                                            scale=True,
+                                            is_training=(mode=='train'),
+                                            updates_collections=None,
+                                            scope=(name+'batch_norm'))
 
     def build_model(self):
         features = self.features
@@ -136,6 +145,10 @@ def build_model(self):
         captions_out = captions[:, 1:]  
         mask = tf.to_float(tf.not_equal(captions_out, self._null))
 
+
+        # batch normalize feature vectors
+        features = self._batch_norm(features, mode='train', name='conv_features')
+
         c, h = self._get_initial_lstm(features=features)
         x = self._word_embedding(inputs=captions_in)
         features_proj = self._project_features(features=features)
@@ -168,6 +181,9 @@ def build_model(self):
     def build_sampler(self, max_len=20):
         features = self.features
 
+        # batch normalize feature vectors
+        features = self._batch_norm(features, mode='test', name='conv_features')
+
         c, h = self._get_initial_lstm(features=features)
         features_proj = self._project_features(features=features)