PDF-to-JSON Professional Extractor

A robust CLI tool for converting PDF documents into structured JSON data. This tool extracts text, layouts, font styles, and images with high precision using PyMuPDF.

✨ Features

• Component-Level Extraction: Extracts text, lines, rectangles, and images as individual JSON components.
• Font Intelligence: Identifies and extracts embedded font styles (Family, Full Name, PostScript) for accurate rendering.
• Visual Accuracy: Captures image positions, dimensions, and transformations.
• Smart Cleanup: Automatically removes overlaps, hidden texts, and redundant background elements.
• Local Storage: Completely standalone with local asset storage (no cloud dependencies).
• Docker Ready: Fully containerized for easy deployment and scaling.

🛠️ Installation

Using Python

1. Clone the repository:

   git clone https://github.com/x-eight/pdf-to-json.git
   cd pdf-to-json

2. Install dependencies:

   pip install -r requirements.txt

Using Docker

docker build -t pdf-to-json .

🚀 Usage

Basic Command

Run the CLI by providing the path to your PDF file:

python src/cli.py path/to/your/document.pdf

Options

• -o, --output-dir: Specify where to save the JSON and extracted assets (default: ./output).

Example with custom output:

python src/cli.py input.pdf --output-dir ./my_results

Running with Docker

docker run -v $(pwd)/output:/app/output pdf-to-json input.pdf

📂 Output Structure

The tool generates a structured output in the designated folder:

output/
├── input.json          # Structured data of the PDF
├── images/             # Extracted portrait and inline images
└── fonts/              # Extracted embedded fonts

📝 Technical Details

This project utilizes:

• PyMuPDF (fitz): For core PDF parsing and SVG extraction.
• Pillow: For image processing and optimization.
• FontTools: For deep inspection of embedded font metadata.
• Font Mapping: Uses src/resources/fonts.json to accurately map embedded and standard fonts (Arial, Calibri, etc.) even when obfuscated.

⚖️ License

This project is licensed under the MIT License.