KaenChan/lightqnet-train

How was lightqnet-dm100.pb trained?

Opened this issue · 0 comments

Hi! Could you please share more information about how the lightqnet-dm100.pb available in your other repository was trained?
I guess it was using the student module mobilenet_v3_small_mini_face_dm100_op1s1_confidence.py, is that right?
Was the mobilenet_v3_small pretrained before the distillation? If so, which was the dataset used?