Image visualization + cropping via ImageToNDArray - improve usability

[AlexDBlack]

Currently, with any of the image visualization steps (draw bounding box, draw points, etc) users need to provide an ImageToNDArray configuration if the original image (that was fed into the net) was cropped.
Cropping can occur for example if the image aspect ratio doesn't match the aspect ratio that the network actually operates on (for example: 1920x1080 --> 300x300).

This is easy to forget or not realize, and it's not particularly user friendly.
What we want: a mechanism in the visualization steps to account for the fact that the original image was cropped, before it was fed into the network. This should be hard to get wrong, and easy to get right, even for newer users and non-experts.