'warpAffine' different behaviour with version 4.12 and 4.13

[DylanMMarques]

Hello,

warpAffine has different behaviour in version 4.12 and 4.13.

import numpy as np
import cv2
input = np.ones((100,100))
pts = np.float32([[1,1],
                  [1,2],
                  [3,4]])
translation_factor = 50
t_cv2 = cv2.getAffineTransform(pts, pts+translation_factor)
out = cv2.warpAffine(
    input,
    t_cv2,
    (200,200),
    flags=cv2.INTER_NEAREST,
    borderMode=cv2.BORDER_CONSTANT,
    borderValue=(0.0,),
)

with:

Package       Version
------------- ---------
numpy         2.2.6
opencv-python 4.12.0.88
pip           24.0
setuptools    79.0.1

Image of out:

with:

Package       Version
------------- ---------
numpy         2.4.2
opencv-python 4.13.0.92
pip           24.0
setuptools    79.0.1

Image of out:

The issue is that in version 4.13, warpAffine translates the image by -translation_factor.