numpy array indicator operation [duplicate]

Question

I want to modify an empty bitmap by given indicators (x and y axis). For every coordinate given by the indicators the value should be raised by one.

So far so good everything seems to work. But if I have some similar indicators in my array of indicators it will only raise the value once.

>>> img
array([[0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0]])

>>> inds
array([[0, 0],
       [3, 4],
       [3, 4]])

Operation:

>>> img[inds[:,1], inds[:,0]] += 1

Result:

>>> img
    array([[1, 0, 0, 0, 0],
           [0, 0, 0, 0, 0],
           [0, 0, 0, 0, 0],
           [0, 0, 0, 0, 0],
           [0, 0, 0, 1, 0]])

Expected result:

>>> img
    array([[1, 0, 0, 0, 0],
           [0, 0, 0, 0, 0],
           [0, 0, 0, 0, 0],
           [0, 0, 0, 0, 0],
           [0, 0, 0, 2, 0]])

Does someone have an idea how to solve this? Preferably a fast approach without the use of loops.

Answer 1

This is one way. Counting algorithm courtesy of @AlexRiley.

For performance implications of relative sizes of img and inds, see @PaulPanzer's answer.

# count occurrences of each row and return array
counts = (inds[:, None] == inds).all(axis=2).sum(axis=1)

# apply indices and counts
img[inds[:,1], inds[:,0]] += counts

print(img)

array([[1, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 2, 0]])

Answer 2

You could use numpy.add.at with a bit of manipulation to get the indices ready.

np.add.at(img, tuple(inds[:, [1, 0]].T), 1)

If you have larger inds arrays, this approach should remain fast... (though Paul Panzer's solution is faster)