Extract handwritten digits from image

See the image at https://imgur.com/uf3eUKI.png. How to extract all the digits from the image. I am able to get only first 3 digits

I stopped right on the third symbol. It looks like the pipe | character.

1 Like

It is digit one (1) . Docsumo and few other online tools easily converts it. Python is not able to.

It was a joke. :slight_smile: I wanted to hint you that you did not provide enough information.

Python itself does not contain OCR and I guess that you did not program the OCR yourself. For anyone to be able to help you, please show the relevant part of your code (ideally as a runnable part). Show the output of the code you are getting and the output you are expecting.

The question will probably be about a certain OCR library, let’s hope that there are people here with experience with that library.

I don’t think that OCR is going to cut it with hand writing; OCR is more about recognizing the consistency of printed text and being able to reconstruct said text into ASCII.

For hand written text, it’s going to involve ‘machine learning’ and a large sample of different hand writing from which to learn.

You can write a Python program to do this but will need to use a library. This Google search has several options. I would use TensorFlow + Keras because that’s what I’m familiar with. You should choose the library that works for you.

You could start your search with Yan Lecun’s LeNet project and see if an easy-to-use neural network is available.

1 Like

I think that the generally accepted definition of OCR includes recognition of hand-written characters and text. IMHO practically every modern OCR method uses some kind of machine-learning.

1 Like

While it is true that handwriting recognition is very difficult, and may sometimes benefit from machine learning, commercial products have offered handwriting recognition of one form or another going back to the 1980s.

If you can’t read a sample of handwriting, it is likely that no OCR system will be able to either. But if the handwriting is neat, then it should be no more difficult for an OCR program to read it than it would be for it to recognise letters in different fonts, typefaces, styles (italic, bold, underlined, plain etc) and sizes.

In other words… the ability of OCR to read handwriting depends on how neat or messy the handwriting is, and how good the OCR system is.