Loss and Accuracy Convergence Graph
这是我使用的模型:
model = tf.keras.applications.ResNet50(
include_top=True,
weights=None,
input_tensor=None,
input_shape=(256,256,3),
pooling=None,
classes=10)
model.compile(optimizer='Adam',loss='categorical_crossentropy', metrics=['accuracy'])
history = model.fit_generator(
train_generator,
steps_per_epoch=487,
epochs=15,
validation_data=validation_generator,
validation_steps=244,
callbacks = callback_list,
verbose=1)
我正在使用批量为32的图像生成器进行训练和验证集 增强对于训练是正确的,但对于验证集是错误的