Hi! I’m trying to learn the correct parameters for an image’s scale and crop. If I want to learn the correct parameters for a crop, that means scale + translation. So would my parametrization be:
[a 0 c
0 e f]
or
[a 0 c
0 a f]
or
[1 0 c
0 1 f] all multiplied by a scalar lambda, where lambda is a parameter?
Thanks!