3 Replies Latest reply on Feb 14, 2018 4:40 AM by MartyG

    RealSense SDK2.0 and hand tracking/gesture recognition algorithms

    veracurly

      Are there any hand tracking and hand gesture recognition algorithms in RealSense SDK2.0?

      Does the D400 camera work with previous version of SDK?

      What camera and SDK version are better if I'm going to work with hand gesture recognition?

        • 1. Re: RealSense SDK2.0 and hand tracking/gesture recognition algorithms
          MartyG

          The previous RealSense SDKs do not work with the 400 Series cameras, unfortunately.  They require the RealSense SDK 2.0.  To gain hand recognition functionality in SDK 2.0, you would need to combine SDK 2.0 with other software platforms such as OpenCV.  Hand tracking modules in OpenCV come in two types, ones that only requires an ordinary webcam rather than a special depth camera like RealSense, and ones that need a depth camera such as Kinect and RealSense.

           

          The link below has details of OpenCV modules that only need an ordinary webcam.

           

          Re: Hand Tracking and Shipping Times

           

          An easier way to achieve hand tracking may be to use an older RealSense camera such as the SR300.  Its '2016 R2' SDK has in-built support for hand detection and tracking.  The USB camera version of the official Intel SR300 is no longer available for purchase (the caseless circuit board module version of it continues to be produced), but there are a couple of third-party alternative USB cameras that are fully SR300-compatible and work fine with the '2016 R2' SDK.

           

          The first of these is the Creative BlasterX Senz3D.  Creative manufactured the official SR300 for Intel, so the BlasterX is practically identical.

           

          BlasterX Senz3D - - Creative Labs (United States)

           

          The other SR300-compatible camera is the Razer Stargazer.

           

          Razer Stargazer Depth-Sensing Webcam

          1 of 1 people found this helpful
          • 2. Re: RealSense SDK2.0 and hand tracking/gesture recognition algorithms
            veracurly

            Thank you for your fast answer!

            I've already had an old version of the camera (F200) and will try to use this one for my purposes. Are there any plans to support and develop hand gesture algorithms in Realsense in the future?

            • 3. Re: RealSense SDK2.0 and hand tracking/gesture recognition algorithms
              MartyG

              You're very welcome.

               

              Your old F200 will work fine with the in-built hand detection and tracking in the '2016 R2' SDK.  I've used an F200 extensively with these myself.  If you need to download '2016 R2', you can launch a direct 1.8 gb download of it in your browser by clicking on the download link below.

               

              http://registrationcenter-download.intel.com/akdlm/irc_nas/vcp/9078/intel_rs_sdk_offline_package_10.0.26.0396.exe

               

              If you have not used your F200 recently and do not currently have the F200 Depth Camera Manager (DCM) program on your computer, you will need to install that first before you install '2016 R2'. The DCM's direct download link is:

               

              https://downloadmirror.intel.com/25044/eng/intel_rs_dcm_f200_1.4.27.52404.exe

               

              With the RealSense SDK 2.0 onwards, Intel are focusing on community sharing of knowledge / code and on open-source integration with third-party platforms such as OpenCV and ROS, with features provided by modules on those platforms.

               

              Intel's near-future SDK 2.0 plans tend to be indicated by the 'coming up' list of features on the download page for the SDK.  New features are continually being added, such as the first iteration of support for the Unity engine in the most recent version. The next major feature on their to-do list is MATLAB support.  No indication of plans for in-built support for hand tracking have been indicated so far.

              2 of 2 people found this helpful