The short and somewhat cheeky answer is that we recognise differences in people's eyes by recognising differences in people's eyes. You don't need to refer to what we have historically designated "sex" to do that.

But here's the longer answer: I'm sure it's true that in the aggregate you can observe some differences in the eye that correlate with sex. But so what? That, along with any other aggregate difference, doesn't actually validate sex as a useful category. The simple fact is that any way you split a population in two, you will see aggregate differences. These differences are then simply used to reify that categorisation as more important and concrete than it really is.

Let me illustrate this with a hypothetical. About half the population requires glasses (or other vision correction), and half doesn't. If we constructed social categories and social roles around these, people would start caring enough to research what the physiological reality correlates with. Is there a difference in athletic ability between glasses-wearers and non-glasses-wearers? Is there a difference in height? and so on and so on. These real physiological differences are then used to reify the social construct, and when someone invents contact lenses, suddenly people go "but these categories are real! look at all this evidence showing how these categories are different!" and so on and so on

but so what? You can split a population in two however you like. Short and tall, glasses or no glasses, male or female. All come along with lots and lots of associated physical and mental differences in aggregate. Why do we think sex matters more than the others? Certainly not because of any physiological differences that actually matter in the modern world. It's socially motivated.