We should first try to provide a route to GET the pose of an image, e.g.:
https://myInstance.ch/api/images/793/pose
This pose should be served using different conventions for
- its position (e.g. different projection systems) and
- its orientation in space (e.g. different ways of expressing the orientation relative to a reference frame
- eventually geometric errors attached to each of the above values