Customized output size for cupyx.scipy.ndimage.affine_transform~

Hi Guys, Thx for open sourced cupy, I was trying to use its builtin api to replace cv2.resize as stated above, but the execution reported ValueError: total size of new array must be unchanged, is the argument output_shape just a placeholder to keep it consistent with scipy API or there are tricks to get it work? THX!

[not522]

Sorry for late reply. I think always zoom is recommended for resizing. Both are very similar but slightly different for interpolation.

a = np.asarray([[1, 3, 5, 7]], dtype=np.float32)
M = np.asarray([[0.5, 0, 0], [0, 1, 0]])
print(cv2.resize(a, dsize=(2, 1)))
print(cv2.warpAffine(a, M, dsize=(2, 1)))

[[2. 6.]]
[[1. 5.]]

[okuta]

Can you send your test code or/and long exception result?