I have an numpy-array with 32 x 32 x 3 pictures with X_train.shape: (32, 32, 3, 73257). However, I would like to have the following array-shape: (73257, 32, 32, 3).
How can I accomplish this?
Add a comment
|
3 Answers
There are two ways to archive this either np.reshape(x, ndims) or np.transpose(x, dims).
For pictures I propose np.transpose(x, dims) which can be applied using
X_train = np.transpose(X_train, (3,0,1,2)).
1 Comment
Weedjo
Thank you, this is correct. np.reshape did destroy the images, np.transpose worked as suggested :-)
np.reshape(X_train, (73257, 32, 32, 3))
https://docs.scipy.org/doc/numpy-1.10.1/reference/generated/numpy.reshape.html
1 Comment
hpaulj
Better check that on a small sample. Transpose is right if you want to preserve images.
I think you want to do a transpose
>>> a = np.array([[1, 2], [3, 4]])
>>> a
array([[1, 2],
[3, 4]])
>>> a.transpose()
array([[1, 3],
[2, 4]])
>>> a.transpose((1, 0))
array([[1, 3],
[2, 4]])
>>> a.transpose(1, 0)
array([[1, 3],
[2, 4]])
