I have seen several APIs that are able to provide some sort of "skeletal data stream" using a Kinect-like device, by extracting skeletal joint information from depth data in a transparent manner from a developer point of view.
Now I am looking for some library that implements a skeletal recognition algorithm (e.g. the one used in MS Kinect SDK, which should be based on "Real-Time Human Pose Recognition in Parts from a Single Depth Image" by Shotton et al. - see attached link) by taking an image file as input.
Does someone have suggestions on where I can find a library for doing this? I have no preferences on the programming language to be used.
Thanks in advance!
http://research.microsoft.com/apps/pubs/?id=145347
I think you can try Berkeley Multimodal Human Action Database (MHAD) for skeletal depth image
There are two databases that can serve as an example Primal Pictures and Visible Body.
Here is a person back in 2010 that was working on this very subject
Maybe he was able to accomplish this task ?
https://channel9.msdn.com/Forums/Coffeehouse/Kinect-open-source-driver-demo/c08f7f4ea226493c97599e2b0115b49d
Here is demo of the Beta 2 in 2011
https://channel9.msdn.com/Series/KinectSDKQuickstarts/Skeletal-Tracking-Fundamentals
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Human-Computer Interaction