The partnership will see Google integrate Movidius’ flagship MA2450 processor chip into its devices. Movidius says the MA2450 is “the only commercial solution on the market today with the performance and power-efficiency to perform complex neural network computations in ultra-compact form factors,” and that makes it particularly appealing to Google’s project: The aim is to use sufficiently sophisticated embedded technology to perform the image recognition functions that normally require communication with an external server. In other words, rather than having to run an image search through remote technology, a smartphone could recognize something—a face, an object—all by itself.
This development can be seen as part of a larger effort on Google’s part to essentially make technology much smarter at recognizing objects and particularly people, with biometrics playing an important role: Last spring, it unveiled FaceNet, an incredibly accurate facial recognition system; and in autumn the company announced major enhancements to its voice analysis technology. While enabling mobile devices to automatically scan and identify the biometrics of everyone they come across will run up against certain privacy issues, it will also represent a significant technological advancement that could offer many benefits, such as helping the blind to identify people they know, for example. The collaboration was only just announced, though, so it may take some time before such results can be produced.
January 28, 2016 – by Alex Perala