This repository hosts plugins for Umi-OCR, an open-source OCR application that supports extensible OCR engines and components. It caters to users seeking to enhance Umi-OCR's capabilities with various OCR technologies, offering offline and cloud-based solutions with different performance and compatibility trade-offs.

How It Works

Plugins are designed as modular components that integrate with Umi-OCR v2 and later. Users download pre-compiled plugin packages from the Releases page and place them in the UmiOCR-data/plugins directory. The repository details several OCR engines, including PaddleOCR-json (optimized for CPU, AVX instruction set required), RapidOCR-json (lightweight, low resource usage), Pix2Text (for math formulas), TesseractOCR (multi-language, good English accuracy), ChineseOCR (lightweight), WechatOCR (offline WeChat OCR), and Mistral AI OCR (cloud-based API).

Quick Start & Requirements

• Installation: Download plugin zip from Releases, extract to UmiOCR-data/plugins.
• Prerequisites:
  • Umi-OCR v2 or later.
  • PaddleOCR-json requires CPU with AVX instruction set (excludes Atom, Itanium, some Celeron/Pentium).
  • Other plugins generally have no special hardware requirements.
• Resources: Plugin sizes vary; PaddleOCR-json is noted as large.
• Links: Releases, Plugin Development, Umi-OCR Main Repo

Highlighted Details

• Offers a diverse range of OCR engines: PaddleOCR, RapidOCR, Pix2Text, Tesseract, ChineseOCR, WechatOCR, and Mistral AI.
• Includes both offline (CPU-bound) and cloud-based (API) OCR solutions.
• PaddleOCR-json plugin supports mkldnn acceleration for enhanced CPU performance.
• TesseractOCR plugin includes a layout analysis model that can improve document parsing accuracy when Umi-OCR's default is disabled.

Maintenance & Community

• The project is part of the larger Umi-OCR ecosystem. Further community and maintenance details would likely be found on the main Umi-OCR repository.

Licensing & Compatibility

• The README does not explicitly state a license for the plugins themselves. The underlying OCR engines have their own licenses (e.g., PaddleOCR, Tesseract). Compatibility for commercial use depends on the specific plugin's underlying OCR technology license.

Limitations & Caveats

• Users are strongly cautioned against downloading the repository's source code directly, emphasizing the need to use pre-compiled releases.
• PaddleOCR-json has specific CPU hardware requirements (AVX instruction set).
• Pix2Text plugin is noted as having a large file size and slower loading times.