Boris Gribkov

@shiyuanyin Did you solve the problem? I see the same behavior with ResNet-18 with ArcLoss and VGG2 training set, I guess that alignment is important

@shiyuanyin I guess I fixed the problem, in case of not aligned vgg2 dataset training is very sensitive to M parameter, try to change

@Bartzi First of all, thanks for your code! Regarding num_chars and num_words in *.npz, I checked synthadd.npz and mjsynth.npz, in both cases num_chars = 1 and num_words = 23. Intuitively...

I see, it's clear now! Thank you!

Dear @Bartzi , sorry to disturb you, another question. According to your paper Localization network try to find and "crop" individual characters, for example FOOTBALL word at the Fig.1. In...

Thank you! using pre-trained weights looks very promising, will try! Also, I was thinking about the image above too, you are right, the recognition task is very simple - license...

Thank you!

Hi @Bartzi Thank you for the good advise, usage of pre-trained Localizer weights helps a lot!  and final accuracy is about 2% better

Does 1by2 mean 0.5 MobileNet in the original paper? It seems that yes according to the number of kernels. I trained a network using your prototxt, final accuracy is close...

Finally I have increased number of kernels twice, this should be original MobileNet, final accuracy 69.5%, getting of over 70% accuracy is still open question.