CODEWITHSUNDEEP
pythontensorflowkeras
AstroBen
AstroBen

Reputation: 863

Keras: How to extract only certain layers from a tensor

I have a 4-D tensor of shape [6,20,30,6] and I would like to execute the keras/tensorflow equivalent of:

new = np.array([old[i,:,:,i] for i in range(6)])

Any help is appreciated!

Upvotes: 1

Views: 499

Answers (2)

AstroBen
AstroBen

Reputation: 863

Thanks to @rvinas for his answer, I was able to cast this in pure keras.

def cc(x):
    return K.backend.stack([x[:,i, :, :, i] for i in range(6)], axis=1)

Then in the keras model definition:

new=L.Lambda(lambda y: cc(y))(old)

Upvotes: 1

rvinas
rvinas

Reputation: 11895

You could expand the dimension of old, use a comprehension list to select the desired slices and concatenate them along the expanded dimension. For example:

import tensorflow as tf
import numpy as np

tensor_shape = (6, 20, 30, 6)
old = np.arange(np.prod(tensor_shape)).reshape(tensor_shape)
new = np.array([old[i, :, :, i] for i in range(6)])

old_ = tf.placeholder(old.dtype, tensor_shape)
new_ = tf.concat([old[None, i, :, :, i] for i in range(6)], axis=0)

with tf.Session() as sess:
    new_tf = sess.run(new_, feed_dict={old_: old})
    assert (new == new_tf).all()

Upvotes: 1

Related Questions