I think your definition of "tracking" is different from Canon's, which seems (based on descriptions of how AF "cases" work) to refer only to tracking across the frame, hence my request for clarification.
- Tracking - select the object, lock it, track it, keep in focus, whatever my framing is
- Servo - constantly refocus on whatever is available under my focus point, no matter, if it makes sense, e.g. two persons heads, and mistakenly focusing on space between the heads, having bg in focus and persons out of focus. Not much useful imo.
But, semantics aside, we are in agreement about on what we both want - i.e. "select the object, lock it, track it, keep in focus, whatever my framing is" - which is basically what happens with the R5 when you select both AF(face) tracking *and* servo AF.
The issue facing the manufacturers is that the camera needs to recognise and register the object under the AF spot. For that to happen, the AF spot/zone needs to be larger than the object/subject within it, for the subject-recognition algorithms to reliably detect it. There will often be instances where the user wishes to track a small part** of the object in question, so the camera then needs to determine which *fragment* of the detected subject that the user wishes to focus on.
Currently technology is focused on detecting eyes, but IMO a lot more attention needs to be paid now to detecting patterns/textures/shapes, which would enable things other than eyes to be "tracked" more reliably.
**e.g. the number plate of a car rather than the vehicle or its driver, or the ear of an animal whose eye is too small to be detected.
Last edited:
Upvote
0