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

How can i recognize name, address handwritten words?

Open MuruganR96 opened this issue 4 years ago • 1 comments

@Breta01 sir, I have followed your instructions, then trained a model for using 5 Handwritten datasets.

I was testing my model in ocr.ipynb It was not extract correctly for particular Name, Address handwritten words. How can i achieve and overcome this issue @Breta01 ?

and one more problem is, non-cursive words it was not concatenate as a word. it is only represent as separate characters.

How do we concatenate separate characters in the non-cursive words?

MuruganR96 avatar Sep 03 '19 07:09 MuruganR96

Can you please be more specific with your problems? If I understan it correctly, you have low accuracy on some words. I don't have any other dataset. You can try to find some other dataset or improve accuracy of trained model.

About the other problem, you probably want to measure distance between letters and connect it based on some limit distance. That would be the easiest solution.

Breta01 avatar Sep 04 '19 19:09 Breta01