Monday, November 18, 2013

Kinect for Windows v2 dev kit unboxing (video)

I just received my Kinect for Windows (K4W) v2 dev kit and opened it for the first time on camera:

This is the alpha version of the hardware that is being sent to K4W developer program members sometime very soon. The final version may differ, and the power supply situation will definitely be smaller.

The K4W team gave the K4W MVPs their v2 dev kits today, November 17th, during the MVP Summit. (Microsoft MVP = Most Valuable Professional, which is an award for people to contribute to various technical communities.) The team is also generously allowing developer program members to share and demo the pre-release sensor hardware and applications publicly without any NDA restrictions.

I'll follow up with another video in a few days after I am back home and have set up the new sensor with my computer.

Note: In the video, I said that the model number for non-USA dev kits would end in 2. There may also be a 3 version. The difference is only which power supply is included. Don't plug in the dev kit to power in the wrong country!


  1. Cool, still waiting for mine *excited* and slightly *scared*. Not sure if bit off too big a piece signing up for the dev program :D

  2. Hi Joshua, I just order 3 Kinect v2. I want to build up a Kinect network on elderly people's house to recognize their daily activities and give them some health recommendations. Could you give me some suggestions? Thank you!

    1. What specifically are you looking for help on?

    2. Based on my topic, I want to know are there similar previous applications that developed by someone to track human activities. I want to know to what level for the state of the art. Say, motion, action, activity, behavior and multiple people's activity. I also want to know how to fuse three Kinect sensors (is it feasible to implement it on a single PC with Windows 8).
      If you have any algorithms suitable for this kind of application, please let me know. Many thanks!

    3. Several people and companies have worked on using Kinect for in-home monitoring for elderly or others. There are also some companies like that do activity analysis, although they are more for retail use, or several companies does motion analysis for physical rehabilitation exercises.

      You can use multiple sensors at once, although the Kinect SDK v2 only allows a single Kinect v2 per computer at the moment. If you want to use three Kinect v2, you would need to use three PCs and share the data via the network.

      For fusing data, it depends upon how accurate the data needs to be lined up, you can use a variety of calibration techniques from opencv stereo calibration with chessboard patterns, to simpler manual data entry calibration. Either way the idea is you are able to transform the data from each Kinect into a global coordinate system. The transformation may involve just translation (if the Kinects are all perfectly lined up and you don't need much accuracy) or full 6 DOF transformation matrices.

      Sorry I don't have any specific links on algorithms for activity or behavior tracking. That type of thing varies so much for each application and requirement. Try searching for research papers on those topics.