handwriting-ocr icon indicating copy to clipboard operation
handwriting-ocr copied to clipboard

How to use my own model to recognize the handwritten text?

Open djk111 opened this issue 6 years ago • 1 comments

Hello,I have trained a model for recognizing a single Chinese handwritten character. The last thing I want to do is identify the handwritten text, which is to upload a handwritten image, and then identify the character. Please tell me what should I do? Can I learn from your project? Thanks a lot!

djk111 avatar Apr 11 '18 03:04 djk111

Hi,

If I understand your problem correctly, you have a model for recognition of individual characters, so you need to process input images. I am not experienced with Chinese handwriting, so I just explain what my code does and what could be useful for you.

In my code, the first step is page detection which removes a background from a photo of a page. Then there is word detection which returns bounding boxes of individual words (if they are separated by space). Maybe you could try to tune the parameters in word detection, so it returns bounding boxes of individual Chines characters. Other steps of my code are normalization and recognition of words which aren't useful for you.

The recognition of Chinese handwriting sounds interesting could you please keep me informed how it goes.

Breta01 avatar Apr 14 '18 17:04 Breta01