README.md

metadata

title: DeepSeek OCR Demo
emoji: 🆘
colorFrom: purple
colorTo: red
sdk: gradio
sdk_version: 5.44.0
app_file: app.py
pinned: false
license: mit
short_description: An interactive demo for the DeepSeek-OCR model.

DeepSeek-OCR Document Recognition

This Space uses the DeepSeek-OCR model for document text recognition and extraction.

Features

• Multiple model size options (Tiny to Large)
• Free OCR and Markdown conversion
• Support for various document types
• Powered by ZeroGPU for efficient inference

Usage

1. Upload an image containing text
2. Select model size (Gundam recommended for documents)
3. Choose task type
4. Click "Process Image"

Model Sizes

• Tiny: 512x512, fastest
• Small: 640x640, good balance
• Base: 1024x1024, high quality
• Large: 1280x1280, best quality
• Gundam: Optimized for documents with crop mode

Credits

Model: deepseek-ai/DeepSeek-OCR