ofxDLib new addon


#1

Hey lovely people,
just pushed an addon for using the dlib library.
It is still a work in progress, by now just implements the dlib face tracker.
It is based on an of app I posted here
It works correctly with projectGenerator, so hopefully it will be easy to integrate to other projects.
Check it and let me know how it works.

Best!

Edit:
sorry I didn’t post the link to the addon. :stuck_out_tongue:


#2

Hey Roy.

Thanks for making this a proper addon.
Works great.
I added an object tracker and a HOG object trainer example to my fork of ofxDLib.

I have nerver sent pull requests. Let me know if you are interested in adding this.


#3

Hey @stephanschulz this is great! I just merged and made some little updates to your code.
If you’re very into this I can grant you access so you dont need to PR instead just push. Just let me know.


#4

dlib version 19.2 is out with neural net face detection + etc.
here is my very first attempt at making it run. it runs very slow and not very good.


#5

Hi @stephanschulz,
Great!
I’m super busy right now so I can’t even try to check it. Maybe I will be able during the weekend.
did you see the following?


All the best!


#6

@stephanschulz thanks for the work, I have been using the hog trainer and get great results. I have a quick question. I thought I could use the trainer and then get the HOG vector and use it to detect the trained object in a video stream. Maybe I am barking up the wrong tree, but if you have any tips on how to save the trained data set and use it later it would be great.

Cheers

Fred


#7

here it says how one would save the trained detector.
http://dlib.net/fhog_object_detector_ex.cpp.html

not sure if i ever tried it.

    // Like everything in dlib, you can save your detector to disk using the
    // serialize() function.
    serialize("face_detector.svm") << detector;
    
    // Then you can recall it using the deserialize() function.
    object_detector<image_scanner_type> detector2;
    deserialize("face_detector.svm") >> detector2;