I feel like we’ve seen this technique used on a crime show or two over the years, but apparently it’s never been something that’s actually possible until recently.
I guess we just all assumed it was coming.
Now, researchers at the University of Maryland have developed a similar technique called neural radiance fields (NeRF). So far, they’ve been able to compute a 3D rendering of a Kirby toy and a classroom scene, all pulled from the reflection in a subject’s eye.
That said, we need to remember these are all created in a controlled environment, and real-world application will be a lot trickier and hard to come by.
Even so, it’s impressive work and it definitely has some exciting potential. It’s also worth noting that it only requires a single, fixed camera and an eyeball.
They were able to approximate using the geometry of the human cornea, which is usually pretty close, and then to measure the number of pixels a cornea takes up in an image before positioning the eye in space.
It’s easier to do this for people who have dark colored irises, as brighter ones can further muddle any reflection.
Which is all to say that real life hasn’t quite lived up to an episode of CSI yet.
But it’s getting there.