Dolphin Document Parser

Dolphin Document Parser converts PDFs and document images into structured content for developers working on document extraction. It handles digitally created and photographed pages, including documents that mix text with tables, formulas, and code.

• Whole-document extraction: Parse document images and multi-page PDFs into JSON and Markdown, or process a directory of documents in batches.
• Layout and reading order: Identify page elements and their reading sequence so extracted content follows the document's intended order.
• Individual element parsing: Extract text, tables, formulas, or code from images of individual elements when you don't need to process a whole page.
• Photographed documents: The model distinguishes photographed pages from digital documents and uses a parsing approach suited to each type.

Dolphin runs locally through Python inference code with downloadable model weights from Hugging Face. It supports Hugging Face Transformers for integration into document-processing applications. The supplied interface consists of scripts for page, element, and layout parsing.