It is worth noting that Facebook is big on both Augmented Reality (AR) and Virtual Reality (VR), both for entertainment and communication purposes. And new research suggests it is working on AR applications that could not just modify or replace your face, but your entire body.
A blog post today has the AI Camera Team showing off a bit of work clearly aimed at doing full-body replacement or tracking in a VR or AR context.
“We recently developed a new technology that can accurately detect body poses and segment a person from their background. Our model is still in research phase at the moment, but it is only a few megabytes, and can run on smart phones in real time,” the researchers write.
Of course, this type of research is far from unprecedented; skeletal tracking systems are commonplace in many industries. And indeed this blog post is more about how this particular system and its component neural networks operate than claiming any major advances.
That said, Facebook is clearly looking especially into keeping things efficient and easy to deploy on mobile. That means working within many hard limitations as far as sensor data, image resolution and refresh rate, processing power available, and so on. Mask R-CNN, as the technique is called, is a nice step in this direction.
We can surely expect more like this from Facebook in the future — the company is even looking for a research intern dedicated to this line of research.