This package provides Optical Character Recognition (OCR) capabilities by leveraging state-of-the-art vision-language models through Ollama. It targets developers and users needing to extract text from images and PDFs, offering both a Python library and a Streamlit web application for flexible integration and use.

How It Works

The core approach utilizes Ollama to serve various vision-language models (LLaVA, Granite3.2-vision, Moondream, Minicpm-v). Users select a model and can process single images or batches, with options for custom prompts, output formats (Markdown, Plain Text, JSON, Structured, Key-Value, Table), and language specification. Image preprocessing is also supported.

Quick Start & Requirements

• Installation: pip install ollama-ocr
• Prerequisites: Ollama must be installed and running. Required models need to be pulled via Ollama (e.g., ollama pull llama3.2-vision:11b).
• Resources: Requires Ollama and downloaded vision models.
• Docs: Ollama OCR on Colab, Example Notebook

Highlighted Details

• Supports PDF and image files.
• Offers multiple output formats including structured data and tables.
• Includes batch processing with parallel workers and progress tracking.
• Provides a Streamlit web application with a drag-and-drop interface.
• Allows custom prompts and language specification for enhanced accuracy.

Maintenance & Community

• No specific contributors, sponsorships, or roadmap details are highlighted in the README.

Licensing & Compatibility

• MIT License. Permissive for commercial use and integration with closed-source projects.

Limitations & Caveats

The README notes that the LLaVA model can sometimes generate incorrect output. Specific performance benchmarks or detailed error handling mechanisms are not provided.