PDF2EPUB is a powerful tool designed to convert PDF files into well-structured Markdown and EPUB formats, leveraging advanced AI-driven layout detection. This project is ideal for users seeking to extract text and images from PDF documents while maintaining the integrity of the original layout.

Features

• Smart Layout Detection: Efficiently handles books and academic papers by accurately identifying and preserving layout structures.
• Advanced Text Extraction and OCR: Capable of extracting text even from scanned documents with Optical Character Recognition (OCR).
• Table Detection and Formatting: Recognizes and formats tables within PDFs for clear presentation in Markdown and EPUB.
• Image Extraction and Optimization: Extracts and optimizes images from PDFs, ensuring quality preservation.
• Clean Markdown Output: Produces Markdown files that maintain the structured content of the original document.
• Customizable EPUB Generation: Allows for tailored styling of EPUB files to suit user preferences.
• Multi-Language Support: Capable of processing documents in various languages seamlessly.
• GPU Acceleration Support: Utilizes NVIDIA and AMD GPU acceleration for improved performance.

Usage Examples

Converting a PDF document is straightforward:

python main.py input.pdf

For batch processing, simply use:

python main.py input_directory/

Advanced command options enable further customization:

python main.py [input_path] [output_path] [options]

With options for batch size, page limits, and language specifications, users can tailor the process to meet specific needs.

Output Structure

The output will be organized into a directory with the following format:

output_directory/
├── document_name/
│   ├── document_name.md
│   ├── document_name.epub
│   ├── document_name_metadata.json
│   └── images/
│       ├── image1.png
│       ├── image2.jpg
│       └── ...

Contribution Guidelines

Contributions to PDF2EPUB are encouraged. Potential contributors should fork the repository, create a new branch, and submit a pull request after implementing changes. A commitment to style consistency and testing is appreciated.

This project benefits from excellent open-source libraries such as marker-pdf, mark2epub, PyTorch for GPU support, and Transformers for comprehensive text processing.

PDF2EPUB is poised to enhance productivity by simplifying the conversion of documents, making a significant impact for users in academia, publishing, and beyond.