This is just a pretty cool way to show your boss that you’ve actually been working.
This visualization was created from one of our old projects SVN repository using Gource.
This is just a pretty cool way to show your boss that you’ve actually been working.
This visualization was created from one of our old projects SVN repository using Gource.
I’ve been offline for over a year in a commercialization effort and am just getting around to posting some results from this period.
This video shows a dual-hand tracking system I developed in OpenCV for desktop applications using a two retro-reflective TrackIR cameras.
The system gave 100fps calibrated 3D position with 1 marker and 6D with 3 or more per hand.
While today there are much more affordable solutions for this purpose but I’ll keep this solution around for cases when accurate fast position tracking without wires is needed. The IR smart cameras are also pretty cool in general but as webcam solutions improve along with depth image sensors like the Kinect, Xtion and PrimeSense cameras, their are fewer and fewer applications where they are the best solution.
Check out the news article about how the beta testing is proceeding.
This is my first attempt at one and two handed camera grabbing using the 3D webcam points.
It’s workable as a demo but not terribly usable with the 3D cursors being invisible before grabbing. This style of interface works best when the cursors are always visible. Having real buttons is to grab and release the world is much more powerful. I had to add nearly critical damping on the data to reduce camera noise. I also added an “up” constraint to keep the world from tumbling out of control.
Most of this was expected, I just wanted to see camera grabbing before I focused on doing the actual “kid friendly” interface which will have a smart game style camera rotating in increments to keep objects in view that are being moved by the same method shown here. Objects with rotation snap constraints in pitch, yaw and scale. Gonna show this next…
Progress is slow, but I’ve now reached a point where the two handed 3d point based interfaces are possible with a reasonably good webcam. This video is using an HP deluxe webcam that delivers 640×480 at 30fps. I’m also cross-testing on a Logitech Quickcam Chat with 320×240 at 15-30fps. While this does work on the Quickcam the difference in FOV and quality is significant.
I’ve added a simple calibration procedure that adjusts for different webcam FOV and operating areas that I’ll document later in a tutorial vid for the kids that need to use the SketchNTell app I’m developing.
There’s also a camera config file now along with an app panel on the tools menubar.