CODEWITHSUNDEEP
tensorflowkeras
user288609
user288609

Reputation: 13055

regarding set_image_dim_ordering('tf') in keras

When experimenting with a Keras implementation, I did as follows:

from keras import backend as K
K.set_image_dim_ordering('tf')

It generates the following error message. What does it mean? I think set_image_dim_ordering is included in Keras.

File "train.py", line 14, in <module>
K.set_image_dim_ordering('tf')

AttributeError: 'module' object has no attribute 'set_image_dim_ordering'

Upvotes: 3

Views: 14908

Answers (8)

Ali_Baghayeri
Ali_Baghayeri

Reputation: 185

Keras is Deep Learning framework which can uses from 'TensorFlow' , 'Theano' and 'CNTK' as a backend. Every backend has its own preferences to use the channel ordering

  • Theano: channel first order
  • Tensorflow: channel last order
  • CNTK: channel last order

if you are using 'Theano' as a backend, you should set "channels_first order" and after import keras library you can use this line of code:

keras.backend.set_image_data_format('channels_first')

and data shape is as follows : (channel, rows, columns).

Note: channel order in your "data shape" should be equal to channel order in the "backend". for example if you are using Tensorflow as backend, then the input shape of your data should be channel last order(for RGB image: (rows, columns, channel))

Upvotes: 5

Shervin Emami
Shervin Emami

Reputation: 2775

To support both new and old versions of Keras, for theano and tensorflow, you can use this:

try:
    if K.backend() == 'theano':
        K.set_image_data_format('channels_first')
    else:
        K.set_image_data_format('channels_last')
except AttributeError:
    if K._BACKEND == 'theano':
        K.set_image_dim_ordering('th')
    else:
        K.set_image_dim_ordering('tf')

(taken from "https://github.com/Arsey/keras-transfer-learning-for-oxford102/blob/master/util.py")

Upvotes: 0

Shaik Ahmad
Shaik Ahmad

Reputation: 569

For tensorflow V2.3

import tensorflow.keras.backend as K
# to know the image order
K.image_data_format()
# to set the image order
K.set_image_data_format()

Upvotes: 0

Biswajit Jena
Biswajit Jena

Reputation: 21

In Keras 2.3.1 uses:

from keras as backend as K
K.common.set_image_dim_ordering('th')

Yes. Its working for me for the Keras version 2.3.1

Upvotes: 2

Hamodey_
Hamodey_

Reputation: 86

What version are you using?

"In the backend, set_image_ordering and image_ordering are now set_data_format and data_format."

Keras 2.0 Release notes

Upvotes: 0

luangrillo
luangrillo

Reputation: 1

In Keras 2.3.1 uses:

from keras as backend as K
K.common.set_image_dim_ordering('th')

Upvotes: 0

Julie G
Julie G

Reputation: 41

Try this :

from keras import backend as K K.tensorflow_backend.set_image_dim_ordering('th')

Upvotes: 2

user2661357
user2661357

Reputation: 5

You might need to do like this:

from keras import backend as K
K.set_image_dim_ordering("th")

Upvotes: -3

Related Questions