Face tracking with dlib?

I have exactly the same MacBook. Will try to debug it. Thanks!

@roymacdonald

i played more with your code and added face feature drawing (eyes, nose, jaw, etc)

i also got the object tracking code working that is shown here:


but i am not sure if it is any better than the flow-keypoint example if ofxCv

here my dlib code:

@stephanschulz Great! I’ll try it and let you know how it goes. :smiley:
The X11 thing is just for windowing. It can be disabled form dlib, but is easier to install it.

@gonzzza it runs quite fast for me too. what’s slowest is the face detection, but once it is detected the face tracking runs super fast and smooth. Do you have any other apps opened while running it? Check the ActivityMonitor (/applications/utilities) and see how much cpu and memory resources is this code consuming as well as to check for other applications using lots of resources.
All the best!

@stephanschulz, Hey I wanted to try your landmark detection code, the objectTracker is great, but as soon as I find a face with

#ifdef USE_FACETRACKER

I get an error in this line

    tempLeft += ofVec2f( d.part(i).x(), d.part(i).y() ); // det.part(i);

in

ofVec2f ofxDLib::getLeftEyeCenter(const full_object_detection& d, unsigned long _faceIndex){

or any of the functions that use d.part(),

Any clues on how I can solve this?

do you have shape_predictor_68_face_landmarks.dat in your data folder?
also m9dfukc have a nice fork on his github

it does landmark smoothing and gives you an object for each found face

Yes I had it, but the fork is great, cheers.