There are some interesting similarities between facial recognition and iris recognition. Scale is not one of them.
Both faces and irises recommend a minimum distance between features. While face distances vary depending upon the algorithm, some recommend a minimum 32 pixels between the eyes. For irises, John Daugman recommended an iris radius of 70 pixels.
Of course, on a human face, the real distance between eyes is much larger than the distance within a small part of an eye.
So what?
If a person’s face is moving rather than stationary, and the camera shutter speed isn’t optimized, the resulting image suffers from motion blur.
Try submitting THAT to a facial recognition system. Garbage in, garbage out.
But it gets worse. What if the system in question were an iris recognition system?
Compare with an excellent quality image with no blur. Yes, this is an AI simulation, but you get the idea.
And that is why irises, despite their impressive accuracy, are not the best solution for all use cases. Which use case is better suited to iris matching?
Bredemarket 