While Tesseract has had some major improvements over the last few years, I still find that it performs much worse than the (paid) computer vision APIs offered by Google, Azure and AWS.
Does anyone have recommendations for other open source OCR engines?
The short answer: nope. It's a bit like speech recognition and synthesis. The open source stuff never matches up to commercial grades services by AWS and Google.
Unrelated, but I recently had to archive some docs and photos and had to start looking for a solid scanner client / OCR software. Prizmo Pro for Mac is incredible, and was one of the only OCR apps I could find that utilized a local RNN.
It’s results are nearly perfect and fast, 3-4x as accurate as adobe acrobat and easily 6-7x the speed.
Even as a dev it blows me away things like OCR running locally on a MacBook has come this far.
Does anyone have recommendations for other open source OCR engines?