CODEWITHSUNDEEP
keras
JMath
JMath

Reputation: 63

how do I get the true labels when I use a ImageDataGenerator.flow_from_directory

I have a code running where as data input I have two numpy array (X_train,y_true). I like the data augmentation of the ImageDataGenerator.

Can I use this for getting corresponding numpy arrays?

Here is some code:

train_data_dir="Path to directory containing for each class a directory of images"

from keras.preprocessing.image import ImageDataGenerator
datagen = ImageDataGenerator(rescale=1. / 255,
                         horizontal_flip=True,
                         rotation_range=360)
generator = datagen.flow_from_directory(
       train_data_dir,
       target_size=(img_width, img_height),
       batch_size=22,
       class_mode=None,
       shuffle=True)

x=generator.next()

Now x a a np.array, containing images of both my classes. Can I find the corresponding array with labels somewhere?

Upvotes: 3

Views: 7968

Answers (1)

Daniel Möller
Daniel Möller

Reputation: 86600

It's quite simple. A generator must output both x and y:

x, y = generator.next()

Another option depending on your python:

x, y = next(generator)

Your generator is not returning any Y, though, because you used class_mode=None.

You should use one of these to make the generator produce labels:

  • categorical
  • binary
  • sparse

Usually, for a multiclass purpose, you'd go with "categorical". For one class (yes/no), use a "binary".

Upvotes: 10

Related Questions