Windows Android Studio numeric OCR using C++

[Michael Jenkins]

I am currently working on a project for an android app that requires rapid OCR on the decimal numeric character set (0 – 9) for digital displays. Ideally it needs to cater for segmented digits as well as standard fonts.

I already have had fairly good success with pre-processing and a custom OCR algorithm for segmented digits. The performance of the current solution is acceptable ranging from 30–100 fps and up. The existing project consists of an android library using OpenCV and custom code that is consumed by the end application. The library is written in C++ using a Windows based Android Studio IDE.

I have the following related questions and would be very grateful for any advice. I would also like to discuss the possibility for a collaboration if anyone is interested.

1) What would you recommend as a good generic, trainable OCR for digital numeric identification?

2) Is there an appropriate library for C++ development within Android Studio using Windows?

3) Which OCR is best for rapid identification of the decimal, numeric character set?

4) I have briefly looked at Tesseract & Leptonica but would ideally like to continue to use C++ in Android Studio on Windows for development and I can’t find a lot of videos or blogs covering this particular setup. Would a Java based implementation in Android Studio be fast enough and could it easily be combined with the existing C++ plug in library?

5) I am also considering Keras and Tensor Flow. Does anyone recommend this option or any other alternative OCR libraries?