CODEWITHSUNDEEP
python-3.xkeras
skoundin
skoundin

Reputation: 192

Keras.utils fails to Import

Below image shows part of the code to train a simple deep CNN(Convolutional Neural Network), on CIFAR small images data set. I have imported, import keras.utils ( Highlighted in red)

enter image description here

However, I still get the below error:

enter image description here

Upvotes: 3

Views: 6721

Answers (3)

Faizan Amin
Faizan Amin

Reputation: 458

Try importing np_utils from keras.utils and use like this

from keras.utils import np_utils
np_utils.to_categorical(y_train, num_classes)

Upvotes: 1

Anurag Sharma
Anurag Sharma

Reputation: 167

You're following a wrong practice of importing python modules. You should go for either of the following practices:

from keras.utils.np_utils import to_categorical

y_train = to_categorical(y_train, num_classes)
y_test = to_categorical(y_test, num_classes)

                   OR

from keras.utils import np_utils

y_train = np_utils.to_categorical(y_train, num_classes)
y_test = np_utils.to_categorical(y_test, num_classes)

You can call only the module/function that you've imported. Say, if you use,
from keras.utils.np_utils import to_categorical
that means you're importing to_categorical function from keras.utils.np_utils package. So, you can only call to_categorical function. But you're trying to call keras.utils.to_categorical, which is not imported. Also, you can't directly import to_categorical from utils without importing np_utils first.

Rule of thumb: If you type from X import Y, that means you've to call Y() as it is and not X.Y(). Doing that is redundant as well as wrong.

Tip: You don't need to mention num_classes as a parameter in to_categorical. Python interpreter will intelligently do that for you.

Upvotes: -1

Sreeram TP
Sreeram TP

Reputation: 11917

You can import to_categorical in keras like shown below.

from keras.utils.np_utils import to_categorical

It can be used like shown below.

Y = [1, 2, 1, 2, 3, 4, 1]

Y = to_categorical(Y)

print(Y)

# output
array([[0., 1., 0., 0., 0.],
       [0., 0., 1., 0., 0.],
       [0., 1., 0., 0., 0.],
       [0., 0., 1., 0., 0.],
       [0., 0., 0., 1., 0.],
       [0., 0., 0., 0., 1.],
       [0., 1., 0., 0., 0.]], dtype=float32)

Upvotes: 5

Related Questions