search

patrikhuber / superviseddescent

C++11 implementation of the supervised descent optimisation method
http://patrikhuber.github.io/superviseddescent/
Apache License 2.0
402 stars 188 forks source link

Question about "data/rcr/face_landmarks_model_rcr_22.bin" #38

Closed xkx0430 closed 7 years ago

xkx0430 commented 7 years ago

Hi Patrik,

I want to ask data/rcr/face_landmarks_model_rcr_22.bin. The algorithm uses only 22 key points to compute the regression and predicts 68 key points. I have tried to train the 68 feature points directly, and the result is not very good. Can you tell me why? Is it because of the regress's regularization parameter lamda?

Thanks a lot.

patrikhuber commented 7 years ago

Hi,

Please see #35, #28 for how I trained my 68 point model. The result is quite good (state of the art in 2014/2015 or so), nowadays there are better models but it's still very usable, and I'm sure it can be made better with more pose training data.

Thanks.

Larumbergera commented 7 years ago

Hi,

Could you tell me which ones are better?

Regards,

Andoni.

patrikhuber commented 7 years ago

I suggest you look at any CVPR 2016 papers or CVPR 2017 submissions. We also submitted one, see here.

Larumbergera commented 7 years ago

Thanks!

xkx0430 commented 7 years ago

Thanks! I got it.