输入的数据用完了； 中断训练

Question

I am trying to train a model with the dataset consisting of 160 images (80 -cars, 80 -planes) using image data-augmentation generators on TensorFlow v2.1. I am getting an error when I run the following code:

classifier.compile(optimizer='adam', loss='binary_crossentropy',
                   metrics=['accuracy'])
from keras.preprocessing.image import ImageDataGenerator
train_imagedata = ImageDataGenerator(rescale=1. / 255, shear_range=0.2,
        zoom_range=0.2, horizontal_flip=True)
test_imagedata = ImageDataGenerator(rescale=1. / 255)
training_set = \
    train_imagedata.flow_from_directory('data/training_set'
        , target_size=(64, 64), batch_size=32, class_mode='binary')
val_set = \
    test_imagedata.flow_from_directory('data/val_set'
        , target_size=(64, 64), batch_size=32, class_mode='binary')
history=classifier.fit(training_set, steps_per_epoch=30, epochs=30,
                         validation_data=val_set,
                         validation_steps=30)

The error is:

Found 160 images belonging to 2 classes.
Found 40 images belonging to 2 classes.
Epoch 1/30
 5/30 [====>.........................] - ETA: 5s - loss: 0.5002 - accuracy: 0.8313WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 900 batches). You may need to use the repeat() function when building your dataset.
WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 30 batches). You may need to use the repeat() function when building your dataset.
 5/30 [====>.........................] - 2s 416ms/step - loss: 0.5002 - accuracy: 0.8313 - val_loss: 1.6599 - val_accuracy: 0.5000

Please suggest what could be done here to rectify this error? Thanks in Advance!

Answer 1

ImageDataGenerator has a default batch_size of 32. you asked to perform 30 steps per epoch, which means 30*32 images per epoch, but you have only 160 images, hence after 5 batches the training collapse. you need to set steps_per_epoch to be floor(num_of_images / batch_size) .