Document understanding: OCR + layout

Documents are not flat images. They have structure: headers, paragraphs, tables, captions, footnotes. Document understanding models combine OCR (text recognition) with layout analysis to extract structured information from PDFs, scans, and forms.