Trained your model on the ETL character data base

[nicill]

Hi, thanks a lot for sharing your code.

I retrained with the data at:

http://etlcdb.db.aist.go.jp/

The results so far look pretty good. Just let me know if you are interested in adding this to your repository.

Also, have you published this code in any format that can be cited in a research paper?

[KyotoSunshine]

Hi nicill, Thanks for the comment. This code has not been officially published, so please just refer to it by URL if you must.

[androbi-com]

Hi nicill, do you have a repo with your trained models? I was thinking about doing exactly that ....

[nicill]

So far I still do not have the models uploaded anywhere, I am actually trying to make a "complete" database of Kanji classes which is not easy.

In any case, if you want to have a pretty decent Handwritten Kanji recognizer, I suggest simply using the files provided at the ETL database to create individual Kanji images (see, for example, http://etlcdb.db.aist.go.jp/?page_id=2461 ) and just substitute the image database that you create inside of this code.

[androbi-com]

Thanks nicill, that is what I am currently doing, but with the code from Charlie Tsai's paper.