One, open source engine
Cuneiform for Linux-originally a Windows software, this is a port of Linux, and it has ceased maintenance in April 2011.
GNU Ocrad-command line tool. There is JS transplantation, which can be used for the front end.
GOCR-command line tool. There is JS transplantation, which can be used for the front end.
Tesseract-open source OCR engine, there are also command line tools. HP developed Google to take over. Support training after 3.0. Golang binding. Getting started tutorial.
OCRopy-OCR engine based on training, after training, it can achieve higher accuracy than Tesseract, and the project is younger than Tesseract. Contains a layout analyzer called OCRopus. in Python.
Microsoft OCR Library —— OCR engine is built-in after Windows 8.1, which can be used on desktop and Windows Phone.
Abbyy-paid software, there is SDK, there is a Cloud version.
Two, preprocessing code
OpenCV-the big brother of image processing. There is Scene Text Detection in OpenCV3 that is worth using.
Libccv-modern image processing library, recommended by many people. Several selected image processing algorithms are implemented, which are clean and easy to transplant. Among them, Stroke Width Transfor is particularly useful.
lswms-branch detection.
OCRopus —— Layout analysis library based on neural learning network algorithm. Tutorial.
TiRG —— text area detection library, effect demonstration.
unpaper —— Detect text and rotation, using Hough transform algorithm.
Three, android mobile terminal
tess-two: Android porting of Tesseract, tutorial.
microblink: Free mobile OCR-SDK.