Passport reader

Demo: https://passport-ear.netlify.app/

Motivation

To not waste time filling out the form where passport data is needed, use a car an excavator that will do it for you

Stack

API: Python, Fastapi, Tensorflow Object Detection Model, Tessecta

Web: React

Mobile: React-Native

How does it work?

The user takes a photo of the main page of the passport that is loaded on the server where the machine-scientific model (Tensorflow Object Detection Model) recognizes the passport fields (determines that the name is the name and surname) and Tessect (OCR) then reads these fields and sends what he reads as JSON. (Reading accuracy is not 100% especially if some fields fall under the shadow or bright light). There is both a web application and a mobile application

Accuracy

Details of fields (Object Detection) of about 100% reading (OCR) about 75%. It confuses some letters, does not recognize some letters. The font is very solved (not all passports are written in one font) and shadows and lights strongly solve

Teaching

The machine model studied at 15,000 passports