tbb can improve face detection speed ?

[antimodular]

cmusatyalab/openface#85 (comment)
have you heard about the CLM-framework (a.k.a Cambridge Face Tracker)?
This post says it can make the dlib face tracker even better.
But i am not sure how much better and if this only works on windows.
I tried to add it to the ofxDLib. After adding CLM-framework-master/lib/3rdParty/tbb/include/tbb to ofxDLib/libs these compile errors:
https://gist.github.com/antimodular/883ce91beef094d7fe5d34250c661823

in general the CLM-framework seems very interesting:
https://www.youtube.com/watch?v=V7rV0uy7heQ&feature=youtu.be

[roymacdonald]

Hey, Nope I haven't heard about this one. I haven't been researching about face tracking for a while now.
About those errors.
it might be that you are not linking the cpp files to your xcode project. make sure that the cpp files from tbb are included in the target project. Like the screenshot below, when you select the cpp on the left pane it must have the highlighted checkbox checked.

To avoid this remember to check the checkbox on the popup that appear when you drag and drop files or folders into xcode. it used to be check by default but it is not for a while now.

also, does the tbb library come with some precompiled stuff? If it does it might be only for 32 bits.

Best

[ofZach]

CLM is really good, I am using it on OSX now w/ dlib and it's great (esp when combined with this CLAHE http://imagej.net/Enhance_Local_Contrast_(CLAHE)) . It requires a ton of additional libraries so I'm not entirely sure how to share it / package it...

[ofZach]

ps: I like it also b/c ofxFaceTracker hates my beard, HoG from dlib and CLM are much more forgiving... used it for a mask project (there's some videos on my ig page https://www.instagram.com/zach.lieberman/)... it tracks great in all kinds of adverse lighting conditions.

[roymacdonald]

Hey @ofZach thanks for these tips. I'll give em a try!
Best!