G[int(wi[1]), int(wi[0])] why flip dimensions?

[pannous]

    # draw estimated gradient as white arrow
    R = np.array([G[int(wi[1]), int(wi[0])] for wi in wp])

why not G[int(wi[0]), int(wi[1])]     (G[xi,yi]) ?

[Andrej]

numpy arrays are accessed as array[y, x].