Inception block as keras layer
Inception blocks should be implemented as true keras layer instead of a method which calls several layers. Call of an inception block currently looks like
inception_model = InceptionModelBase()
X_in = Input(...)
X_out = inception_model.inception_block(X_in, conv_settings_dict1, pool_settings_dict1,
regularizer=self.regularizer,
batch_normalisation=True,
padding=self.padding)
but should become
X_out = inception_model.inception_block(conv_settings_dict1, pool_settings_dict1,
regularizer=self.regularizer,
batch_normalisation=True,
padding=self.padding)(X_in)