This project extracts structured outlines like Title, H1, H2, H3 from PDF files using an ML model trained offline — as part of Adobe Hackathon Round 1A.

• Extracted font and layout-based features using PyMuPDF
• Converted JSON ground truth + PDFs into a dataset (CSV)
• Trained an XGBoost classifier using TF-IDF vectorized features
• Evaluated using precision/recall metrics
• Saved model, vectorizer, and encoder
• Configured everything to run in Docker (offline)

docker build --platform linux/amd64 -t adobe-headings:v1 .