LightOCRX: A lightweight C++ implementation of PaddleOCR

Hello everyone,

I'm working on a project called LightOCRX. It's a full Paddle OCR pipeline (Detection, Classification, Recognition) written in C++ with haru pdf C library.

Most OCR solutions today are heavy unportable non reproducible Python programs with several hundred megabytes of dependencies. I wanted something portable, lightweight, easy on system resources and fast to compile.

Features:

* ONNX Runtime for inference using Paddle ONNX models.

* Using stb_image to keep the program tiny (no OpenCV), and clipper library to handle Paddle text detection logic.

* Searchable PDF, text only PDF or plaintext export.

* Handles rotated text correction and deskewing for images (mostly) using paddle classifier.

Link: https://github.com/IvanRolero/LightOCRX

Any feedback is welcome.

https://redd.it/1s7f86q
@r_cpp