Tensorflow for digitizing handwritten and printed documents?

Jacques_Vaessen · February 23, 2022, 3:56am

Hi, first of all I am not sure if this is the right forum for my question/queste…
I work at DigitalOcean. One of my customers is the Dutch National Archive who is in need of a proven technology to digitize their archives.

I read the following tutorial [How To Build a Neural Network to Recognize Handwritten Digits with TensorFlow | DigitalOcean] and was intrigued.
Could Tensorflow be the technology for digitizing old (handwritten and printed) documents?

Would be great if someone could point me in the right direction.
Best,

Jacques

lgusm · February 23, 2022, 11:57am

Hi Jacques,

Yes, you can build an OCR with TensorFlow
Here are some resources to help you:

Model to give you a good start from TFHub: Find Pre-trained Models | Kaggle . This might already solve your problem.
If you want to build something yourself, here’s a tutorial: OCR model for reading Captchas

Jacques_Vaessen · February 23, 2022, 12:51pm

Hi Luiz,

Thank you for reaching out to me, much appreciated. Being an AE, I am from the dark sales side and I lack the technical knowledge to dig deeper in the TF product offering, so it is hard to know what exactly I can propose to the end-customer.
Is there a sales (engineer) team that could get involved in my conversation with the customer to advise on the proper way forward?

Best regards,

Jacques

Bhack · February 23, 2022, 5:15pm

I suggest you to take a minimal overview on this domain:

Bhack · February 24, 2022, 1:46am

You could also be inspired by some quite vertical in production services like:

https://calfa.fr/

lgusm · February 28, 2022, 12:51pm

TensorFlow team does not have a sales engineer (as far as I know).
If your client is willing to do the ML themselves, you can direct them here too to get help when needed.