Transfer Learning 을 할때에 BatchNormalization layer를 학습시키는지 아닌지에 따라서 결과차이가 차이남 from tensorflow.keras import layers f1 = tfa.metrics.F1Score(num_classes = CLASSES, average='macro') x = tf.keras.layers.Input(shape=(400,400,3), name = "image") base_model = tf.keras.applications.EfficientNetB3( include_top=False, weights='imagenet',input_tensor = x) base_model.summary() for layer in base_model.layers: ..