PDF Image Copyright Analyzer

A Python tool that analyzes images in PDF documents for copyright considerations using Claude AI. The tool processes each image and:

• Classifies them into copyright categories
• Finds potential source citations through reverse image search
• Generates alternative images for replaceable content
• Produces a comprehensive report

Classification Categories

Images are classified into one of four categories:

• N: Not copyrighted/Creative Commons/Public Domain
• P: Permission needed (academic/university/federal/non-profit content)
• F: Fair use applicable
• R: Needs recreation/replacement

Features

• Extracts and analyzes all images from PDF files
• Performs reverse image search using Google Images
• Attempts to extract citations from source websites
• Annotates the original PDF with classification labels
• Generates a JSON report with detailed classifications
• Uses Claude 3.5 Sonnet for image analysis
• Automatically generates alternative images using Stable Diffusion (optional)

Requirements

• Python 3.x
• Chrome WebDriver (for Selenium)
• API Keys:
  • Anthropic API key (for Claude)
  • Stable Diffusion API key (for image generation)
  • 2captcha API key (for solving captchas - optional)

Installation

pip install anthropic python-dotenv PyMuPDF stability-sdk selenium beautifulsoup4 trafilatura 2captcha-python

Environment Setup

Create a .env file with your API keys:

CLAUDE_API_KEY=your_claude_api_key
STABILITY_API_KEY=your_stability_api_key
CAPTCHA_API_KEY=your_2captcha_api_key  # Optional

Usage

Basic usage:

python main.py path/to/your.pdf

With alternative image generation:

python main.py path/to/your.pdf --generate-alternatives

Show browser during citation search (helpful for debugging):

python main.py path/to/your.pdf --show-browser

Output Files

1. Annotated PDF (*_annotated.pdf):

   • Original PDF with classification labels (N/P/F/R) next to each image

2. Classification Report (*_classifications.json):

   • Detailed information for each image
   • Includes classifications, citations, source URLs
   • Alternative image paths (if generated)

3. Alternative Images (optional):

   • Generated images for content marked as 'R' or 'N'
   • Saved as separate PNG files

Known Issues and Tips

1. Citation Search:

   • Using --show-browser flag helps monitor the citation search process
   • Useful for debugging when citations aren't being found correctly
   • Allows visual confirmation of successful Google Image searches

2. Captcha Handling (Work in Progress):

   • Current implementation of captcha solving has limitations
   • The URL being sent to 2captcha API may not match the actual captcha site
   • Manual intervention might be needed for sites with strict captcha protection

3. Performance Considerations:

   • Processing large PDFs with many images may take significant time
   • Image search and text extraction are the most time-consuming operations