Greetings.
I've trained an OCR model to populate simple data (first name, surname, dob, id number) and the identified patterns are all correct. However, when adding new data, the model does not work correctly: for example, it also extracts the "date of birth:" expression and not just the actual date (eg. 2023-05-06); also, even tough it correctly distinguished first name vs. surname during training, it sometimes extracts the full name into each section (that is, both "first name" and "surname" columns are populated with the full name).
Could someone help me tweaking this model to achieve a better performance?
Also, I understand the OCR model is currently in its beta version; its been a while in this state (at least 3 years). Should we expect any improvements in the near future?
Thank you for your time.
LinkedIn
Twitter
