AI Claude Document Parser is an intelligent tool designed to extract information from various types of identification documents, such as passports and ID cards. Using advanced image processing and AI-powered text recognition, this tool can parse important details from document images and compile them into a structured format.

Intellectual Authors: Dawar Ali & Ernesto Cohnen

For more information, visit ixigo.tech

• Supports multiple image formats (PNG, JPG, JPEG, WebP) and PDF files
• Extracts key information such as document type, personal details, and issuance information
• Utilizes Anthropic's Claude AI for accurate data extraction
• Processes entire folders of documents in batch
• Outputs results in CSV format for easy analysis and integration
• Implements image resizing and compression to optimize API usage

graph TD
A[Input: Image/PDF] --> B[Resize & Compress]
B --> C[Encode to Base64]
C --> D[Send to Claude AI]
D --> E[Parse JSON Response]
E --> F[Save to CSV]
G[Process Next File] --> B
F --> G

1. Clone this repository:

   git clone https://github.com/ixigo/document-parser.git
   cd document-parser
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Set up your Anthropic API key as an environment variable:

   export ANTHROPIC_API_KEY='your-api-key-here'
   

Run the script with the required input and output parameters:

python documents_parser.py --input /path/to/your/images --output /path/to/your/output.csv

The --input parameter specifies the folder containing the images to process, and the --output parameter specifies the path for the CSV output file.

The script generates a CSV file with extracted information. Here's an example of what the output might look like:

| filename | documentType | country | passportNumber | surname | givenName | dateOfBirth | gender | placeOfBirth | placeOfIssue | dateOfIssue | dateOfExpiry |
|----------|--------------|---------|----------------|---------|-----------|-------------|--------|--------------|--------------|-------------|--------------|
| passport1.jpg | Passport | United States | 123456789 | Doe | John | 01/01/1980 | M | New York | New York | 01/01/2015 | 01/01/2025 |
| id_card.png | ID Card | Canada | AB123456 | Smith | Jane | 15/05/1992 | F | Toronto | Ontario | 10/10/2018 | 10/10/2028 |

The script includes robust error handling:

• If a file cannot be processed, an error message is displayed, and the script continues with the next file.
• If the AI response cannot be parsed as JSON, the script attempts to extract JSON from the response text.
• In case of API errors or other exceptions, detailed error messages are logged.

The script optimizes image processing to meet API requirements:

• Images are resized to a maximum of 2000x2000 pixels.
• Images are compressed to stay under 5MB while maintaining acceptable quality.

Contributions are welcome! Please feel free to submit a Pull Request.

This project is licensed under the MIT License - see the LICENSE file for details.