While we haven’t tested the Z9’s Eye AF accuracy yet, we’ve often found that Nikon’s Eye AF (Z7/II, recent firmware) is:
- Desire to find eyes/faces in a scene most, leading to many false positives
- Tends to focus first.
We surmise that Nikon’s machine learning-based approach is more loosely trained, with less stringent criteria.
Sony/Canon is conservative but has fewer false positives.
With the Z9, though, things have progressed considerably. It can recognize faces/eyes in much smaller scenes than Sony or Canon, although this is in line with the general trend that Nikon’s Eye AF is trained to over-recognize eyes/faces rather than recognizing faces. rheumatic (with the risk of multiple false positives). As for focus position accuracy (front-to-back focus) as well as false positives/negatives, we’ll have to wait to comment on that until we can further test the Z9, R3 and A1- next to.
Suffice it to say that all 3 cameras are phenomenal and highly effective.
