Originally posted by derekkite Six months is an eternity when applied to finishing and optimizing.
It depend what you call optimizing.
AF for example there basic algorithms for sure and you just need to implement them.
But then you need to have sensor that work in dim light. And you need to have many, to cover the frame and have an accurate vision of the scene. Even that can be done fast, ok. But don't forget, you also need AF point that work well too at f/8 f/9 for slow tele with TC as well as f/1.4 f/2 for very precise focus. And it must work in daylight, artificial light etc.
Now you want to have tracking. A good part of this is pattern recogition on the metering sensor to see what moved. This is subject still under heavy research in general because nobody managed to have a machine that really see. Even google with thousand of computer in working on the task and billions of sample image is far from being there.
Now you also want predictive AF. The basic idea is likely very simple to implement if you do it. Predict the speed of the subject and predict where it will be exactly at the moment you take the photo while the AF sensor can't see it. But then again to reduce a precise trajectory of something that is basically a bunch of pixels is not that easy.
All of this need to be tuned, optimized and tested against many tries and use cases. This is really research and you could spend the next 10 years doing only that and not just you but your whole team. Just setting up the protocol to test and evaluate the results are already a complex topic.
No, 6 months is not a lot.