Why don't we change the intrinsic matrix scale?

[Zhong-Zi-Zeng]

Hello,

This work is very impressive. However, I have a little problem.

When you use the get_geometry function, I noticed that although the feature map is smaller than the original input images, you didn't change the intrinsic matrix.

Does this cause any problems? If you can answer this for me, I would greatly appreciate it. Thanks!