I want to contribute a code for facial landmark detection and head pose estimation to opencv
Hello community,
I have worked on facial landmark detection and head pose estimation, and have a working solution for it. I want it to be a part of open source community i.e. opencv.
Method used is supervised in nature (more particularly cascade of linear regressors). During training, I used some fast algebra library (armadillo on top of lapack and blas) for faster calculation of least squares. I used ibug-300 faces in the wild dataset (which consist of AFW, LFPW, HELEN, ibug databases) for training, and model seems to be performing fairly well.
Now, I have a trained model, and I can run the code without any external dependency apart from opencv. So, can I make it as a part of opencv community (I am referring to only detection part using pre-trained model that I trained, not training part) ??
oh, that's nice. maybe you could try to add it here ? (have a look at the face module, which already has similar implementations)
https://github.com/opencv/opencv/wiki...
just out of curiosity:
how large is the model ? would it be possible to reuse existing code ? do you have a github repo, we can look at ?
Hi, Thanks for the info. I do not have any git repo yet, but I am planning to make one. Model file is around 80 Mb.
please do so. (you need some github skills for this anyway...)
and keep a close eye on the face module (tests, coding style, etc.)
any updates ?
Hi berak, I made a git repo that contains an executable and source code for the same. Check it out and let me know your feedback.
that's cool !
made a little writeup there, have a look, please.