Open issac8huxley opened 8 years ago
@issac8huxley No preprocessing is needed. AFLW data is enough strong to converge and easy to train 2 class classification. You may check your data whether there are wrong labeled picture or too hard. The final error on training data of 12net is about 5%
I test your training code on the AFLW dataset, but the 12-net could not convergence. Would I make some pre-processing on the train data? Thanks