I had a similar problem, trying to recognize digits in natural scene photographs and getting about 5% accuracy for digit crops. I ended up using "train-bpnet-isolated" and "rec-bpnet-isolated" scripts, which had the effect of producing a digit-only classifier, and got accuracy of 33.4% on 6.5k digit dataset
I'm considering trying the 0.4 training architecture on the dataset next, and I'm checking this group regularly for updates on 0.4 documentation http://yaroslavvb.com/upload/strands/examples/digits.html On May 15, 3:58 am, Pierpaolo Monaco <[email protected]> wrote: > Thanks for your replay Tom. > I'll try to do that. > > ----------------------------- > Pierpaolo Monaco > ---------------------------- --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "ocropus" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/ocropus?hl=en -~----------~----~----~----~------~----~------~--~---
