Layout detection and OCR models that turn patent PDFs and images into structured, reading-order content.