YangUSTC

model can't converge

16

Hi there, I'm trying to use picolrn to train my own face detectors (using AFLW dataset). first i use genki.py to preprocess my positive samples(24384faces),then i choose around 15G background...