biubug6/Pytorch_Retinaface

About the ratioHard Example Mining

mawj111 opened this issue · 3 comments

mawj111 commented
About the ratioHard Example Mining
mawj111 commented

@biubug6 Hello! After reading the code, I have a question about the ratio of HEM. I see that the ratio of positive and negative samples is set to 1:3 in SSD, why do you set the ratio of positive and negative samples to 1:7? I would appreciate it if you could reply. Thank you very much.

mawj111 commented

Pytorch_Retinaface/train.py

Line 82 in b984b4b

criterion = MultiBoxLoss(num_classes, 0.35, True, 0, True, 7, 0.35, False)

mawj111 commented

@biubug6 Also, why set threshold to 0.35 rather than 0.5? I would appreciate it if you could reply.