search

jwyang / face-alignment

Face alignment in 3000 FPS
322 stars 190 forks source link

21 landmark dataset #27

Open yujinfu opened 8 years ago

yujinfu commented 8 years ago

Hi, could you please tell me how to train a model in 21 landmark , download a 21 landmark dataset or just use a 68 landmark dataset and change the global parameter landmark number to 21?

jwyang commented 8 years ago

hi, you may do not need to download a 21 landmarks dataset, you can alternatively train the model with 68 landmarks by setting the params.ind_usedpts to those indice of your 21 landmarks. Please refer to line 10 in config_tr.m:

params.ind_usedpts = 1:68; % [1:17, 28, 31, 32, 34, 36, 37, 40, 43, 46, 49, 52, 55, 58];

yujinfu commented 8 years ago

Hi, it's me again I saw a video in youtube which show the powerful face alignment in 3000 fps ,the link is that: https://www.youtube.com/watch?v=TOVFOYrXdIQ I wonder how to set the initial parameters and train the model to make it work like the video show. I use c++ code from the link https://github.com/yulequan/face-alignment-in-3000fps thanks for your help.