ruiming46zrm
ruiming46zrm
> ut name. The new version has been changed. I have already modified before
change 1: {"cls_pred_stride_8", "dis_pred_stride_8", 8}, {"cls_pred_stride_16", "dis_pred_stride_16", 16}, {"cls_pred_stride_32", "dis_pred_stride_32", 32}, change2: customer information like num_class any other to change?
set lr = 0.1 at the begining.
halo, any tricks?
> > halo, any tricks? > > No tricks, only increased the batchsize and learning rate to 128 and 1e-1 respectively. your cfp-fp performance is quite good~ ! better than...
in the "face_verify.py" , you firstly should make the arg.update == true , then run this py , in the facebank you will see facebank.pth , after then , reset...
by modified network , large batch_size = 384 (4 GPUs), lr = 0.1 , milestone = [3,6,9,12]，drop ratio = 0.4 . train ir_se50, get : lfw accrancy : 99.78% agedb-30...
@bnulihaixia as above : 1. shortcut; 2:add bn . you may try to use large batch and large lr to get better result
@twmht can you tell us your acc on cfp-fp by res50, mine is 87.5%