In the image plane here we want to add XYZ axis (red, green, blue):

Note that the camera already has axes, but not the image. The camera space != image space (the difference is a pinhole projection).

Secondly, we want to use the view coordinates to label all XYZ axes everywhere (if available).

So add labels to the XYZ axes in:

• 3D space
• The visualization of rigid3 transforms in a 3D space (usually cameras)
• The image plane projected in 3D space (the image above).
• The 2D view (X & Y only)

Lastly, the 2D view should take view coordinates into account so that you can have Y=up, for instance. That is: the 2D viewport should flip/rotate images that are not using X=right, Y=down. Currently the 2D viewport always assumes X=right, Y=down.