Numpy Array replace booleans with string depending on position of "True"

To process to output of a multi-class classification I’d like to process a numpy array in such a way, that every True from the first column results in a class1 and a True in class2 correspondingly. A row with no True‘s should be translated into class3.

My initial array looks like that:

[[True False],
[False False],
[False True],
...
[True False],
[False True]]

(A row containing [True True] can not arise.)

What I’d like to get out is:

[class1 class3 class2 ... class1 class2]

Ideas for an elegant and fast approach are highly appreciated. Thanks in advance!

>Solution :

You can use numpy.select.

import numpy as np
cls = np.array([[True, False],[False, False],[False, True],[True, False],[False, True]])

mask = cls.any(-1)
condlist = [(mask & cls[..., 0]), 
            (mask & cls[..., 1]),
            (mask==False)]
choicelist = ['class1', 'class2', 'class3']
res = np.select(condlist, choicelist, 'Not valid')
print(res)

Output:

['class1' 'class3' 'class2' 'class1' 'class2']