Skip to content

HoloLens 2.0 Will Use A New Processing Unit To Implement Speech And Image Recognition

HoloLens 2.0 Will Use A New Processing Unit To Implement Speech And Image Recognition

The first version of Microsoft’s HoloLens Mixed Reality headset has plenty of room for improvement, but the company is working on a new version that will sport a new processing unit to enable the device to use deep learning.

Marc Pollefeys, Director of Science on the HoloLens project, recently said as much in a new Microsoft blog post. He revealed that the company is upgrading the Holographic Processing Unit (HPU), which handles all of the information from the kits onboard cameras, sensors, and its inertial measurement unit (IMU). This upgraded unit will enable the use of Deep Neural Networks (DNNs) via an AI coprocessor, which might allow the device to recognize anything from objects to voice patterns and more.

Harry Shum, EVP of Microsoft AI and Research, presented the upgraded HPU at the CVPR 2017 conference over the weekend with a live demo of hand segmentation. Everything is done on-device instead of needing to send information to the cloud, reducing the latency in between actions.

Most excitingly, the blog confirms that this new HPU will be seen in the next version of HoloLens, which might be the first time we’ve heard them acknowledge another iteration is in the works (though it was kind of obvious). Imagine being able to fully operate the next iteration of the device with your voice, perhaps interacting with an AI or something like the company’s Cortana virtual companion. That’s just the start of where the tech could be taken.

Right now HoloLens is a $3,000 development kit; will that be the case for the next device? What other upgrades can we expect? We’ve got plenty of questions, but that’s nothing new, is it?

UploadVR Member Takes

Weekly Newsletter

See More