mtcnn-caffe

[jYaog]

Hello,your another project close the 'issue ' fuction. The project name is ‘mtcnn-caffe’ . The problem is that the code doesn't create '12/pts.imdb' ,but the layer that you created has used the imdb. thanks for your help.

[994088895]

I want to know if 12net uses 12.pts, or it only used 48.pts in 48net?

[boyang24]

@CongWeilin @jYaog because the mtcnn caffe do not have the issues,so i can only do it as this way, the steps introduction is not that clear,after run gen 12 net output the negative part positive data,and i use two gpus to run hard examples,it takes more than 10 hours,anyway this is ok,but after that,i found the steps is wrong,we must to run gen 24net data.py then run hard examples,it is ok,you tell us to check the Seanlinx,yes i did,but 12net needs the pts,24net need the pts,where we find the pts,you did not give the method to make the pts by celeb database,so how to do it?

[jhgfkdj]

@boyang24 hi, have you solved this problem? I confused the same as you did. Have you got a useful get_net_imdb.py? I am used to using tensorflow and a new caffe learner.