I searched previous topics but they are quite old. I am hoping there are some good news.
The best I found is a guy that managed to import some Java library in a Mac, - but I am on windows and have no idea what to do to import that library.
It is 4 years old.
What I am trying to do is to look at printed labels with a USB camera.
When in focus, do a freeze frame and press a button that will generate a string that contains the characters of the OCR. I know how to use cameras and make a image from video.
Then I would generate a new row in a CSV for each recognition, I know how to do that too.
All labels are black on white, but the font type and size varies.
My boss gave me about a thousand labels to that I am now processing by hand, like in medieval times.
Please help, I hate data entry,
If you want to try to get OCR working in Processing for fun or for an art project, definitely go for it.
However, if you are trying to get work done efficiently, this might be a bad idea. Default Tesseract – even with good focus and stable text sizes – is not great, and it is extremely susceptible to problems with lighting, resolution, focus, skew etc. You will probably get significantly better performance from a commercial desktop OCR package or even a cheap or free iOS or Android OCR app, or from an OCR appliance (like a pen scanner or portable book scanner, depending on the size and form factor of what you are scanning). So much so that, if you hate data entry, home-rolled OCR probably isn’t even worth considering as an option, as you may end up having to review and edit a huge number of scans.
I am headed down this road myself. It is too bad to see that there is not more support for this.
I am comparing between using Python, Unity, and Processing. Would be great to use Processing more for OCR but with not many libraries and most of the replies saying “don’t try it” ^^ it is easy to see why people don’t.
I will update a larger forum posts with any exploits and updates.
Does OCR work differently to a neural network?
I have found that a really good route is to get a document camera that has OCR functionality. Get one that can save to a .txt file - like the ICODIS x7. Boom, up and running!