jackdack

Hey guys, first post on here! I'm creating a project in which I need to detect whether someone is looking left or right relative to the laptop screen. I've adapted a colour tracking patch from this forum and also designed a bespoke pair of glasses with 3 unique colours on (the patch tracks three individual colours) to try and accomplish this, however it's kinda temperamental and takes a TONNE of calibration so I'm hoping someone on here can help with a better, more functional patch. I've been advised to use a kinnect 'hack' but that means buying a kinnect; I'm trying to be as frugal as I can which is why I want to use the built-in camera.

Any suggestions would be greatly appreciated!

Hey guys, first post on here! I'm creating a project in which I need to detect whether someone is looking left or right relative to the laptop screen.  I've adapted a colour tracking patch from this forum and also designed a bespoke pair of glasses with 3 unique colours on (the patch tracks three individual colours) to try and accomplish this, however it's kinda temperamental and takes a TONNE of calibration so I'm hoping someone on here can help with a better, more functional patch.  I've been advised to use a kinnect 'hack' but that means buying a kinnect; I'm trying to be as frugal as I can which is why I want to use the built-in camera.
Any suggestions would be greatly appreciated!
J.


180˚-face-recognition-using-macbrook-pro-camera-can-it-be-done

A quick search turns up this: http://www.tomhall.com.au/blog/?p=3604


Ok, downloaded and ran this option. I have the cv.jit library, and similarly this FaceOSC has about a 60˚ range from centre (30˚ either side of the camera) as it uses algorithms to detect facial features. When someone is looking 90˚ (hard left or hard right) their face doesn't show 'normal' traits that these algorithms look for so they don't work. They tend to need to see both halves of the face. Please, keep the ideas coming!

Ok, downloaded and ran this option.  I have the cv.jit library, and similarly this FaceOSC has about a 60˚ range from centre (30˚ either side of the camera) as it uses algorithms to detect facial features.  When someone is looking 90˚ (hard left or hard right) their face doesn't show 'normal' traits that these algorithms look for so they don't work.  They tend to need to see both halves of the face. Please, keep the ideas coming!


https://cycling74.com/wiki/index.php?title=Jitter_Recipes

180˚ face recognition using macbrook pro camera. Can it be done?