mxnet and ocr

[wanghaisheng]

we have a lot of jpg file which contain chinese characters and tables and so on.
can we use mxnet to get an ocr model to identify these words and characters in a new picture?
could someone give a little advice

[piiswrong]

Which OCR model do you want to use?

[wanghaisheng]

i have read a presentation here 链接: http://pan.baidu.com/s/1kUoSyV1 密码: p92n
they claim to use CNN to identify chinese character and RNN to english characters.

[futurely]

There is a Torch 7 implementation which can be ported to MXNet. keras-team/keras#383 (comment)

[wanghaisheng]

@futurely thx