Exclude BN gamma and beta from L2 regularization.

model.py

@@ -1955,8 +1955,10 @@ class MaskRCNN():
             self.keras_model.add_loss(tf.reduce_mean(layer.output, keep_dims=True))
 
         # Add L2 Regularization
+        # Skip gamma and beta weights of batch normalization layers.
         reg_losses = [keras.regularizers.l2(self.config.WEIGHT_DECAY)(w) / tf.cast(tf.size(w), tf.float32)
-                      for w in self.keras_model.trainable_weights]
+                      for w in self.keras_model.trainable_weights
+                      if 'gamma' not in w.name and 'beta' not in w.name]
         self.keras_model.add_loss(tf.add_n(reg_losses))
 
         # Compile