People often doubt that we can deliver what a string of eyecare professionals and multi-national billion-dollar companies have failed to do. Our claims are not bluster, they're an evolution of the science of fitting glasses.  In this first part of a 3-part series, we describe how we can build a highly accurate 3D model of your face.  In part 2, we will describe glasses design, and in part 3 we discuss one-off manufacturing.

It starts with a selfie

A quick video selfie allows us to catch hundreds of high-quality pictures of your face from different perspectives.  It is the different perspectives that gives us the 3D info we need to build a 3D model of your face.  The first thing our computer vision algorithm does is to identify recognizable landmarks in each image, like the tip of your nose or the outside edge of your right eye.  By triangulating overlapping points across multiple images, we create a 3-dimensional map of the surface of your face, containing 8,000 individual measurements that describe what makes  It's similar to how your two eyes perceive the world from two slightly different perspectives, and your brain is able to combine said differences into depth perception.  In our case, all of this is done in a little over a minute of extremely-heavy computation.

But where's the camera?

Astute 8th grade geometry students and other computer vision scientists usually stop us around here and ask about solving camera position, or scale, or objective reference.  What they want to know is: how can we tell the difference between someone who actually has a really big head, and someone who's just holding the camera close to their face so that it LOOKS like they have a really big head? It's actually a great question.

The answer is we can't... yet.  And we don't need to...yet.

The highly precise 3D mesh of your face is very precise on the slopes of each curve, the angles at each junction, and the relative sizes of any set of measurements, but we don't know units of measurement quite yet.  But because we're previewing a custom product, rather than a stock product, that's actually ok.  After you design and order your glasses, the last step of completing your order is to submit a scale recording.  In this video taken with the higher-resolution rear camera, we rebuild the 3D model, this time including an object of known size and shape (a simple credit card stripe), which tells us the true scale of your face model and the glasses you ordered, so we can size them to within a fraction of a millimeter.

You might ask why we need to rebuild the credit card in 3D? It turns out people are really bad at getting perfectly head-on selfies.  If the card stripe isn't perfectly parallel to your face, and perfectly aligned to the axis of measurement, you will mis-measure every time.  Our 3D measurement is far more accurate, repeatable, and reproducible than other inferior online measurement tools that rely on one picture only.

Did this answer your question?