验证损失和准确度都在波动。训练准确度和损失已经饱和

Question

我正在尝试使用带有 2 个类的 mobilnetv2 来训练分类器。我已经冻结了基础层并添加了以下几行。

 headModel= baseModel.output
 headModel = AveragePooling2D(pool_size=(7, 7))(headModel)
 headModel = Flatten(name="flatten")(headModel)
 headModel = Dense(128, activation="relu")(headModel)
 headModel = Dropout(0.5)(headModel)
 headModel = Dense(2, activation="softmax")(headModel)