📝 Quick Summary:

PDF Craft is a Python library designed for converting PDF files, especially scanned books, into other formats like Markdown and EPUB. It utilizes OCR, layout analysis, and optionally LLMs to extract text, remove irrelevant elements, determine reading order, correct errors, and structure the content for improved readability and semantic coherence.

🔑 Key Takeaways

✅ Automates the conversion of scanned PDFs to Markdown and EPUB formats.
✅ Utilizes a combination of AI models (DocLayout-YOLO, OnnxOCR, layoutreader) for accurate text extraction and formatting.
✅ Handles multi-page documents and intelligently preserves reading order.
✅ For larger books, leverages LLMs to create structured EPUBs with chapters and tables of contents.
✅ Open-source and community-driven, with opportunities for contributions and improvements

📊 Project Statistics

⭐ Stars: 2764
🍴 Forks: 160
❗ Open Issues: 24

🛠 Tech Stack

✅ Python

Ever wished you could magically transform those unwieldy scanned PDF books into clean, easily readable digital text? PDF Craft is here to make that wish a reality! This amazing GitHub project tackles the challenge of converting scanned PDFs – think old textbooks or research papers – into formats like Markdown and EPUB, and it does so with a clever blend of AI and some serious coding magic. Forget painstaking manual transcription; PDF Craft automates the entire process.

At its core, PDF Craft uses a multi-stage approach. First, it employs DocLayout-YOLO to pinpoint the text regions on each page, intelligently filtering out distracting elements like headers, footers, and page numbers. Think of it as a super-powered highlighter that only selects the important text. Next, OnnxOCR steps in to recognize the actual text within those highlighted areas. This is where the AI really shines, accurately converting images of text into machine-readable characters.

But it doesn't stop there! PDF Craft goes the extra mile by using layoutreader to determine the most natural reading order, ensuring that the final output flows smoothly and logically. This is crucial for maintaining the integrity of the original document's structure. For smaller documents, the output is clean Markdown, perfect for quick editing and sharing.

For larger books (think 100+ pages), PDF Craft leverages the power of Large Language Models (LLMs) to generate a structured EPUB file. This is where things get really exciting. The LLM not only helps organize the text into chapters and add a table of contents, but it also intelligently corrects OCR errors, resulting in a much cleaner and more accurate final product. Imagine having a smart assistant meticulously proofreading and formatting your book for you. While this step requires an external LLM service (DeepSeek is recommended), the results are well worth the effort.

PDF Craft is a game-changer for anyone who works with scanned documents. Developers can integrate this project into their workflows to automate the conversion of PDFs, saving countless hours of manual work. The ability to generate both Markdown and EPUB formats provides flexibility for various downstream applications. The project's open-source nature allows for community contributions and improvements, ensuring its continued evolution and enhancement. Plus, the use of locally executable AI models reduces reliance on cloud services, providing greater control and potentially lower costs. Whether you're a researcher, a student, or simply someone who hates manual data entry, PDF Craft is a must-have tool in your arsenal.

📚 Learn More

View the Project on GitHub

Enjoyed this project? Get a daily dose of awesome open-source discoveries by following GitHub Open Source on Telegram! 🎉

PDF Craft: Your New Best Friend for Scanned PDF Conversion