Different calibration results python vs cpp

edit

Hi,

I am using the calibration samples in:

1 - Python

2 - C++

3 - Custom python script.

I tried to make sure that all parameters are the same (checkboard layout, stopping criteria, and cornerSubPix mask size. But still the three cases give different results. 1 Is completely wrong because it gives too big focal distances and after rectifying the images, straight lines look bent. 2 and 3 provide similar results, but not the same. In fact, 2 provides the exact same values for fx and fy, which I assume is not likely to happen.

Anybody experienced this before? Any tip on what could be wrong? I am using OpenCV 3.1.0