DesktopVR using OGRE and OpenCV

This is a version of DesktopVR based on Johnny Lee's demo. Its using a standard webcam to do the headtracking. It is using the OGRE 3D engine for visuals and OpenCV for the tracking.

Category:

Science & Technology

Tags:

• DesktopVR
• 01

[proantix]

This is awesome, are you releasing the source code. I need something like this in my PhD research where I'll have a 3D animated character instead of targets. I'm very impressed.

[cybereality]

@proantix I released the source code on the Ogre forums. Just go to Ogre3d (dot) org and goto the forums. Then do a search for "desktopvr". Should be the first thing to come up. I'll admit the code is a little hackish and certainly not PhD quality but it might be something worth toying with to give you some ideas.

[Amaroqdricaldari]

I believe FaceAPI provides better performance. It is a simple Headtracking software that works with a standard camera. It tracks on the X, Y, and Z Axis, and it can handle very rapid head movements over a large interaction space. It can even handle Contortions and Obstructions of the face. Check out the videos by Torbensko.

[sam4san]

(contdd from previous comment)

and draws a view that is a balance between the projected movement and the actual movement.

3) Is there any chance of 'almost' eradicating the faults using 2 cameras?

In any case.. this is a revolution in the making :).. thumbs up dude

[sam4san]

Dude !! awesomely done!. Can I give you a few suggestions? (If I may :D). The little jerks that you have in the viewport probably can be avoided if you ignore sudden user movement - This is based on the assumption that a *normal* human being cannot suddenly jump from one place to other in an instant.

2) May be you could write a 'user movement projection' which kind of predicts the user's next movement based on the previous few seconds or milliseconds movements (contdd...)

[rewb0rn]

Looks nice already, keep up the good work!

[proantix]

@cybereality Thanks.

[sdpagent]

Sorry i need to ammend my comment... using directshow i was able to gain access to the webcam properties and disable lots of things, this increased camera frame speed for both opencv and directshow (both now 15 for my camera) however, i was able to seperate the threads using 'videoInput' library, which means that its faster in the sense that my gameloop can run fast, (getting 60fps) even though my camera is only doing 15, If your good at threading you can do same for OpenCV. Maybe this will help

[sdpagent]

It might not be the facial recognition making it jittery, it might just be the opencv access to the frames and the camera itself. I was using opencv at first as well. Discovered that by using direct show and disabling the cameras light boost and color boost properties, i was able to get over 50% more fps out of the camera (15 instead of 7-8. Mind you I was using a shitty £5 logitech camera though.

[HectorUch123]

Hey really nice! good ideia , see PROJECT MOSSORO , i think you will like. Congratulations!