It’s amazing to think about how far we’ve come in reading and digitizing text!
Press enter or click to view image in full size
Not so long ago, “OCR” meant some clunky software struggling to parse basic fonts from scanned pages. I remember working on a first project some years back (this does not make me feel younger 😆) with Tesseract and OpenCV, and we were happy when we managed to get some Latin characters recognized. We considered it high-tech if it could transform a scanned business card into text without too many errors.
[If you do not have a Medium Subscription — You can read here for free]
You can still find some old projects leveraging such architecture, like here for example: https://github.com/gali8/Tesseract-OCR-iOS
I remember when we migrated from Tesseract to Google Vision SDK, it was already a great step forward, despite latency being a challenge.
A Brief Look Back: The Early Days of OCR
While we often associate OCR with modern software libraries, its roots stretch much further back. Believe it or not, the very first commercial OCR systems date to the 1950s, where bulky mechanical scanners would identify characters by matching them against pre-stored templates. It was a far cry from what…