Recently, we’ve been prototyping some sporting and retail applications built around Microsoft’s Hololens. The work has helped highlight the capabilities and limitations of the technology, one of which is it’s processing power considering it’s enormous potential.
So it’s pretty exciting to hear that Microsoft will be adding on-board machine-learning capabilities to their next Hololens release featuring their HPU 2.0. The HPU – ‘Holographic Processing Unit‘ – is Microsoft’s custom multiprocessor built for the Hololens. This includes object- and facial-recognition and what will presumably be a lightweight neural network or AI processor. It appears that most of this will be achieved without making a bunch of calls to the cloud, rapidly reducing processing time and hopefully creating an experience unhampered by immersion-breaking latency.
Seeing AI (iPhone app)
Microsoft released the Seeing AI app this year, and it’s a good example of the types of things HPU 2.0 will be able to achieve. It’s not (readily) available to us in the Australian app store unfortunately, but try imagining replacing your phone with a headset and you start to get the picture.
The Seeing AI app is intended to aid visually-impaired people to understand their environment in greater detail:
The next Hololens is a little while off, but in the meantime we’ll be exploring the vast possibilities of the latest version.