Dmytro Mishkin

@Juddd thank you! Could you please make PR here? https://github.com/kornia/tutorials/

> @ducha-aiki did the local tests pass ? Yes, but I have forgot that I locally have torch 1.10, where safe_solve works. So let's forget about this PR until we...

Added new test for CI https://github.com/kornia/kornia/actions/runs/3089642255/jobs/4997486509 @shijianjian could you please take a look at `mypy` errors? They are in `kornia.augmentation` module

@shijianjian ping?

Just in case: you can use recently added kornia functions [`camtoworld_graphics_to_vision_Rt`](https://github.com/kornia/kornia/blob/master/kornia/geometry/conversions.py#L1268) and [`camtoworld_to_world_Rt`](https://github.com/kornia/kornia/blob/master/kornia/geometry/conversions.py#L1354) We have even added Apple ARKit pose to colmap conversion: [`ARKitQTVecs_to_ColmapQTVecs`](https://github.com/kornia/kornia/blob/master/kornia/geometry/conversions.py#L1416)

Thank you. One more question - both models (scale and ori) also output other thing - ori and scale. So, what it is best _single_ model? :)

Thank you. And finally - have you checked if the performance degrades on the grayscale images?

Also add add_padding() and remove_padding() functions

Maybe in future, but figuring out now where to put it, would result in many cyclic dependencies between perspective, pinhole, etc camera. I am also not a fan of using...

Also, I am against putting conversions into liegroups without keeping some alias from the more general module. Reason - in many cases user have no idea wtf is the Lie...