📄 PDF Outline Extraction System

🏆 Adobe India Hackathon Challenge - "Connecting the Dots"

🎯 Overview

A smart solution for Round 1A of the Adobe India Hackathon that automatically extracts structured outlines from PDF documents! 📚

✨ What it does: Converts messy PDF documents into clean, hierarchical outlines with headings (H1, H2, H3) and their page numbers in beautiful JSON format.

🚀 Key Features

• 📖 Processes PDFs up to 50 pages quickly
• 🏷️ Extracts document titles and heading hierarchy (H1, H2, H3) with page numbers
• 📊 Outputs structured JSON with page references in the required format
• ⚡ Lightning fast processing (≤10 seconds for 50 pages)
• 🔌 Works offline - no internet needed, no GPU required
• 🌍 Multilingual support (including Japanese 🇯🇵 and other non-Latin scripts)
• 🐳 Dockerized for easy deployment on AMD64 (x86_64) CPUs

📥 Input & 📤 Output

📥 Input

PDF files placed in /app/input directory

📤 Output

Clean JSON files in /app/output with this structure:

{
  "title": "Document Title",
  "outline": [
    { "level": "H1", "text": "Chapter 1: Introduction", "page": 1 },
    { "level": "H2", "text": "Overview", "page": 1 },
    { "level": "H3", "text": "Historical Background", "page": 2 }
  ]
}

🏃♂️ Quick Start Guide

🔨 Build the Container

docker build --platform linux/amd64 -t pdf-outline-extractor:latest .

🎯 Run the Extraction

docker run --rm \
  -v $(pwd)/input:/app/input \
  -v $(pwd)/output:/app/output \
  --network none \
  pdf-outline-extractor:latest

🧠 Smart Algorithm

📋 Title Extraction

🎯 Finds the largest font text on the first page as the document title

🏷️ Heading Classification

Uses intelligent multi-factor detection:

• H1: ≥18pt AND bold 💪
• H2: ≥15pt AND (bold OR UPPERCASE) 📢
• H3: ≥11pt OR numbered patterns (e.g., "1.", "2.1.") 🔢

🌍 Multilingual Detection

Uses langdetect to support documents in various languages, including Japanese native script! 🇯🇵

🧹 Smart Filtering

Automatically removes duplicates and very short text for cleaner results

✅ Technical Requirements (All Met!)

Requirement	Status
🖥️ AMD64 architecture compatible	✅
🚫 CPU-only (no GPU needed)	✅
⚡ ≤10 seconds for 50-page PDFs	✅
🔌 No internet required	✅
📦 Model size ≤200MB	✅
🪶 Lightweight solution	✅

📦 Dependencies

Package	Purpose
🔧 PyMuPDF (fitz)	PDF processing powerhouse
🌍 langdetect	Multilingual language detection
🐍 Python 3.12	Runtime environment
📚 Standard libraries	os, json, re

📁 Project Structure

📦 pdf-outline-extractor/
├── 🐍 main.py          # Core extraction magic ✨
├── 🐳 Dockerfile       # Container setup
├── 📥 input/           # Your PDF files go here
├── 📤 output/          # Generated JSON files
└── 📖 README.md        # This documentation

🧪 Sample Test Results

✅ Successfully tested on:

• 📚 Academic reports with formal structure
• 🏛️ Government forms with simple layouts
• 🔬 Technical documents with multiple heading levels
• 🌍 Multilingual documents (Japanese 🇯🇵, French 🇫🇷, Hindi 🇮🇳, etc.)

🎖️ Challenge Compliance

Criteria	Implementation
⚡ Performance	Fast, lightweight processing
🎯 Accuracy	Multi-factor heading detection (font size, boldness, numbering)
📋 Format	Perfect JSON structure as specified
🔄 Generalization	Works across document types and languages
🚫 No hardcoding	No file-specific logic or web/API calls
🔌 Offline ready	All dependencies included, zero internet needed

🏆 Built with ❤️ for Adobe India Hackathon 2025 - Round 1A

Ready to connect the dots in your PDF documents! 🎯✨