Handwritten Digits OCR

This repository consists of codes for a handwritten digits recognizer. It consists of a paper region extraction module and a digits recognizer. The application is written in c++, with minimum dependency required.

For technical details, visit this blog post.

Usage (Compiled for macOS Mojave)

Place the trained model under model/

./bin/scan input_image output_image output_txt

Dependency

• CImg Version 1.6.2 (included)
  • To load PNG: libpng
  • To load JPG: libjpeg
• Eigen Linear Algebra Library Version 3.3 (included)
• LibSVM for SVM classifier (included)
• Libtensorflow for CNN classifier

Folder Structure

• src code files for OCR software
• model trained model (SVM and CNN) for digits recognizer
• include required library (excluding Libtensorflow)
• bin compiled binaries