Hi Debayan, On Wed, Jun 04, 2014 at 01:53:54PM +0530, Debayan Banerjee wrote: > I am contemplating porting the classifier to a deep neural net, probably > https: > //github.com/BVLC/caffe. Anyone already working on this? > > This should allow Tesseract to recognise some of the more complicated scripts > as well (Indic, Arabic, Chinese).
Tesseract does have a neural net implementation, called 'cube', and it is used for Arabic and a few other scripts. That said, it is now easier to add different classifier systems to Tesseract; see some of the slides and code from Ray's recent talks: https://drive.google.com/folderview?id=0B7l10Bj_LprhQnpSRkpGMGV2eE0&usp=sharing So by all means have a go and see if you can do better than cube! Also there are no training tools available for cube, which is a pain, so a neural net implementation that included them would be a good thing to have. Nick -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/20140604131150.GA8694%40manta.lan. For more options, visit https://groups.google.com/d/optout.
