README.md

---
title: MonkeyOCRv2 Document Parser
emoji: 📄
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 6.15.1
app_file: app.py
short_description: Parse document images into structured Markdown
python_version: "3.12"
startup_duration_timeout: 30m
---

# MonkeyOCRv2 Document Parser

This Space demonstrates **MonkeyOCRv2-B-Parsing**, a visual-text foundation model for Document AI.
Upload a document image and the model will:

1. **Detect layout** — identify text blocks, tables, formulas, pictures, headers, footers, and their reading order
2. **Recognize content** — extract text, render tables as HTML, formulas as LaTeX, and pictures as embedded images
3. **Output structured Markdown** — a clean, readable representation of the document

## Model

- **HF Model**: [`zenosai/MonkeyOCRv2-B-Parsing`](https://huggingface.co/zenosai/MonkeyOCRv2-B-Parsing)
- **Paper**: [MonkeyOCRv2: A Visual-Text Foundation Model for Document AI](https://huggingface.co/papers/2607.11562)
- **GitHub**: [Yuliang-Liu/MonkeyOCRv2](https://github.com/Yuliang-Liu/MonkeyOCRv2)

## Supported Document Types

Receipts, academic papers, forms, tables, mathematical formulas, multi-language documents
(ZH, EN, AR, DE, ES, FR, HI, ID, IT, JP, KO, NL, PT, RU, TH, VI).