What is the right way to rewrite this code?

Question

C++ / C#

Good afternoon. There is the following code.
int win_size = 100;
const string test_image = "C:\\Users\\Roman\\Desktop\\group.bmp";
Mat test_im = imread(test_image, 0);
for (int i = 0; i < test_im.rows-200; i = i + 50)
{
for (int j = 0; j < test_im.cols-200; j = j + 50)
{
Mat testHog = hog(test_im (Range(i, i + win_size), Range(j, j + win_size)));
testHog.convertTo(testHog, CV_32FC1);
int response = svm->predict(testHog);
if (response == 1)
{
rectangle(test_im, Point(i, j), Point(i + win_size, j + win_size), Scalar(), 2);
}
}
}
imshow("win", test_im);
waitKey();
Implemented the hog method, now I'm trying to look for a person in a sliding window. To test, I took 5 small images, which the hog recognized as people during testing, collected them on a white background at intervals between themselves and gave them to the sliding window. As a result: 1) for some reason, on a rectangular image, the sliding window runs only over a square approximately equal to the height of the image 2) does not select people where they are and selects a white background in some places
Tell me what is wrong with my code