PDF to Image Conversion and AWS Textract Analysis

A Python utility for converting PDF documents to images and analyzing expense documents using AWS Textract's AnalyzeExpense API.

Prerequisites

• Python
• AWS Account with appropriate permissions
• AWS credentials (Access Key ID and Secret Access Key)
• S3 bucket for document storage

Installation

# Install required packages
pip install pdf2image
pip install PyMuPDF
pip install amazon-textract-textractor

Part 1: PDF to Image Conversion

Required Libraries

import os
import fitz  # PyMuPDF

Usage

# Configure paths
pdf_file = "<File Path of your pdf document>"
output_dir = "<File path to save the image of your pdf document>"

# Convert PDF to images
pdf_to_images(pdf_file, output_dir)

Function Details

def pdf_to_images(pdf_file, output_dir):
    """
    Converts a PDF file to multiple PNG images (one per page).
    
    Parameters:
        pdf_file (str): Path to the input PDF file
        output_dir (str): Directory where the output images will be saved
        
    Output:
        Creates PNG files named 'page_1.png', 'page_2.png', etc. in the output directory
    """

Part 2: AWS Textract Analysis

Required Libraries

from textractor import Textractor
from textractor.data.constants import (
    AnalyzeExpenseFields, 
    AnalyzeExpenseFieldsGroup, 
    AnalyzeExpenseLineItemFields
)

AWS Configuration

# Set AWS credentials
os.environ["AWS_ACCESS_KEY_ID"] = "<Your AWS Access Key ID>"
os.environ["AWS_SECRET_ACCESS_KEY"] = "<Your AWS Secret Access Key>"

# Initialize Textractor
extractor = Textractor(region_name="ap-southeast-1")  # Replace with your AWS region

Analyzing Expenses

# Analyze expense document
document = extractor.analyze_expense(
    file_source="<path_to_image>",
    save_image=True
)

# Visualize results
document.visualize(with_words=False)

Accessing Analysis Results

1. Access expense document:

expense_doc = document.expense_documents[0]

2. View available data:

# View summary fields
expense_doc.summary_fields

# View summary groups
expense_doc.summary_groups

# View line items groups
expense_doc.line_items_groups

# Convert line items to pandas DataFrame
expense_doc.line_items_groups[0].to_pandas()

Output Data Structure

The analysis provides:

• Summary fields (key-value pairs from the document)
• Summary groups (grouped related fields)
• Line item groups (tabular data from the document)
• Metadata including:
  • Page count
  • Word count
  • Line count
  • Number of tables
  • Number of forms
  • Number of key-value pairs

Important Notes

1. AWS Textract Limitations:

   • Only works with images and documents stored in S3 buckets
   • Region-specific availability
   • Requires proper AWS permissions

2. PDF Conversion:

   • Creates separate PNG files for each page
   • Maintains original resolution
   • Requires sufficient disk space for output files