Remove unnecesary shape from BN layers

edufonseca · Dec 6, 2018 · e1a4622 · e1a4622
1 parent d720c2b
commit e1a4622
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 9 deletions.
diff --git a/architectures.py b/architectures.py
@@ -22,38 +22,38 @@ def get_model_baseline(params_learn=None, params_extract=None):
     spec_x = spec_start
 
     # l1
-    spec_x = BatchNormalization(axis=1, input_shape=input_shape)(spec_x)
+    spec_x = BatchNormalization(axis=1)(spec_x)
     spec_x = Activation('relu')(spec_x)
 
     spec_x = Conv2D(24, (5, 5),
                     padding='same',  # fmap has same size as input
                     kernel_initializer='he_normal',
                     data_format='channels_first')(spec_x)
-    spec_x = BatchNormalization(axis=1, input_shape=input_shape)(spec_x)
+    spec_x = BatchNormalization(axis=1)(spec_x)
     spec_x = Activation('relu')(spec_x)
     spec_x = MaxPooling2D(pool_size=(4, 2), data_format="channels_first")(spec_x)
 
     # l2
-    spec_x = BatchNormalization(axis=1, input_shape=input_shape)(spec_x)
+    spec_x = BatchNormalization(axis=1)(spec_x)
     spec_x = Activation('relu')(spec_x)
 
     spec_x = Conv2D(48, (5, 5),
                     padding='same',  # fmap has same size as input
                     kernel_initializer='he_normal',
                     data_format='channels_first')(spec_x)
-    spec_x = BatchNormalization(axis=1, input_shape=input_shape)(spec_x)
+    spec_x = BatchNormalization(axis=1)(spec_x)
     spec_x = Activation('relu')(spec_x)
     spec_x = MaxPooling2D(pool_size=(4, 2), data_format="channels_first")(spec_x)
 
     # l3
-    spec_x = BatchNormalization(axis=1, input_shape=input_shape)(spec_x)
+    spec_x = BatchNormalization(axis=1)(spec_x)
     spec_x = Activation('relu')(spec_x)
 
     spec_x = Conv2D(48, (5, 5),
                     padding='same',  # fmap has same size as input
                     kernel_initializer='he_normal',
                     data_format='channels_first')(spec_x)
-    spec_x = BatchNormalization(axis=1, input_shape=input_shape)(spec_x)
+    spec_x = BatchNormalization(axis=1)(spec_x)
     spec_x = Activation('relu')(spec_x)
 
     spec_x = Flatten()(spec_x)

diff --git a/main.py b/main.py
@@ -174,9 +174,6 @@
 filelist_audio_tr_nV_small_dur = train_csv.nV_small_dur.values.tolist()
 idx_nV_small_dur = [i for i, x in enumerate(filelist_audio_tr_nV_small_dur) if x == 1]
 
-filelist_audio_tr_nV_small_clips = train_csv.nV_small_clips.values.tolist()
-idx_nV_small_clips= [i for i, x in enumerate(filelist_audio_tr_nV_small_clips) if x == 1]
-
 
 # create dict with ground truth mapping with labels:
 # -key: path to wav