Microsoft just put on the transparent face masks, lined up the sensors and released an updated version of the Kinect SDK. To the high-pitched squeals of (some) developer’s delight, Kinect for Windows SDK 1.7, along with an updated developer toolkit and new Human Interface Guidelines (HIG), is available on the Kinect for Windows website. It’s been in the works for a while with the new Fusion functionality bringing the highly prized 3D scanning capability that renders up those 3D scenes in real time, but are they too late?
3D Scanning with Kinect
We saw the first taunting of ‘KinectFusion’ from Microsoft Research in August 2011. That’s… just a dang long time ago. Of course, many have already been using Kinect Xbox 360 devices to create their own hadheld 3D scanner, with others like Matterport developing an entire new scanning device using the same sensor technology. The Kinect for Windows version is said to be a better commercial solution than the Xbox 360 Kinect with support for “near mode” (recognizing objects as close as 40cm) and API improvements all bent on bringing 3D scanning via the Kinect to the professional.
From the Kinect for Windows Team:
These 3-D images can then be used to enhance countless real-world scenarios, including augmented reality, 3-D printing, interior and industrial design, and body scanning for things such as improved clothes shopping experiences and better-fitting orthotics. Kinect Fusion is something many of our partners have been asking for and we’re thrilled to be delivering it now.
Some of these partners include Boeing, Coca-Cola, Citibank, General Electric Healthcare Group, Konami, Mattel, Telefonica, and Nissan. Microsoft is really pushing the ability to “reconstruct high-quality, three-dimensional (3-D) renderings of people and objects in real time” and they plan to do it in real time via the GPU with MATLAB and OPENCV examples included in the toolkit. The feature list includes:
Real-time, GPU-assisted 3-D object and scene reconstruction by using the Kinect for Windows sensor
- Ability to infer relative sensor position and orientation from a 3-D scene for augmented reality application
Advanced algorithms that are powerful enough for large sensor movements and scene changes during scanning
- DirectX11 compatible graphics cards supported
AMD Radeon 7950 and NVidia GTX560 have been validated to run at interactive rates
Kinect Fusion Studio and samples demonstrate 3-D scanning capabilities
- Non-real time CPU mode for non-interactive rate scenarios
Late to the game?
At this point, Kinect scanning capability still isn’t plug-n-play, but it does give developers an inside track to creating desktop scanning apps and it’s becoming more obvious that Microsoft is interested in extending the reach of the device into industry, working with companies at an enterprise level. But are they late to the game? It certainly looks that way on the consumer level. Xbox 360 scanning is still a hacker industry and with the LEAP Motion Controller shipping in May and their own SDK already in the hands of developers for the better part of a year, Microsoft may have put too much dependence on name recognition and corporate relationships.