Navigation
Recherche
|
Microsoft puts mixed reality, high-speed 3D rendering, and Kinect vision into cloud
lundi 25 février 2019, 18:19 , par Ars Technica
Enlarge / The Azure Kinect Development Kit. (credit: Microsoft)
While HoloLens 2 is undoubtedly the aspirational star of Microsoft's augmented-reality (AR) offerings, the company isn't putting all its eggs in that particular basket. Alongside the new HoloLens headset, the company also announced the Azure Kinect development kit: a new version of the Kinect sensor technology. Though Kinect first shipped as a gaming peripheral, the accessory immediately piqued the interest of researchers and engineers who were attracted to its affordable depth sensing and skeleton tracking. While Kinect is no longer a going concern for the Xbox, the same technology is what gives HoloLens its view of the world and is now available as a device purpose-built for developing scientific and industrial uses for the technology. The Azure Kinect Developer Kit (DK) bumps up the specs substantially when compared to the old Xbox accessory; it includes a 12MP RGB camera, 1MP depth camera, and 360-degree microphone array made up of seven microphones. It also contains an accelerometer and gyroscope. Early customers are using it for applications like detecting when hospital patients fall over, automating the unloading of pallets in warehouses, and creating software that compares CAD models to the physical parts built from those models. The original Kinect used only local compute power; indeed, one of the many criticisms made of it was that the Xbox devoted a certain amount of processing power to handling Kinect data. The new hardware still depends on local computation, with things like skeleton tracking handled in software on the PC the hardware is connected to. But the Azure name is not just there for fun; Microsoft views the Kinect as a natural counterpart to many Azure Cognitive Services, using the Kinect to provide data that's then used to train machine-learning systems, feed image- and text-recognition services, perform speech recognition, and so on. Read 3 remaining paragraphs | Comments
https://arstechnica.com/?p=1463961
|
56 sources (32 en français)
Date Actuelle
ven. 22 nov. - 19:40 CET
|