ofxCcv and face detection

@kylemcdonald i have used ofxCcv 32bit version and of_v0.9.0RC1_osx_release successfully. It’s great.
It is so strange when it identifies things correctly and so funny when it gets it wrong.

I see that here libccv they talk about face detection. is that already possible with your implementation?


i’ve only experimented with the object recognition, and haven’t tried face detection yet. but i’d love to see it added to ofxCcv!

do you think to added to ofxCcv one would have to rebuild the addon to work with the ccv 0.7 or it’s just a matter of teasing out the right data from the current version?

I managed to get the face tracking example code working.
It finds relatively small faces too but, runs at 2fps when used on every update cycle.

There is probably a way to use the face finding only once and after that do something like key-point tracking.
ofxFaceTracker also seems to have fps drop until it finds and tracks a face.

It’s possible too that i might not been using the GPU. Not sure how to check that.


the next thing i want to try is getting the icfdetect function to run as mentioned here:

if i understand this correctly i would find pedestrians:

just don’t know yet how to tab in to the lib.

got the pedestrian tracker working.

as with the face tracker, frame rate is super low right now.


do you know if it is possible to enable GPU (CUDA) usage for the ccvlib? I am specifically hoping to try it for my face tracking code i made for ofxccv.

it’s possible to use CUDA with ccv. it’s significantly more complicated to get set up, but i did it once. it should give you a big speed increase (~5-10x) on deep learning applications.

if you feel like the pedestrian and face detection stuff could be merged with the main oxfCcv branch i’d love to see it included! i can definitely imagine using the pedestrian tracker myself.

ok posted a pull request.

would be interesting to see if that 5-10x increase would also work for the trackers.
do you by any chance have a guide on what to do?