DeepSeek-OCR WebUI

语言: English | 中文

A web-based interface for DeepSeek-OCR model.

📋 Release Notes

[2025-10-24] - Enhanced Table Recognition & Visual Annotation

• ✨ Table Recognition Display: Added enhanced table structure visualization with markdown formatting
• 🎯 Image Position Annotation: Implemented interactive HTML-based coordinate annotation system
  • Color-coded bounding boxes for detected text regions
  • Real-time coordinate legend with precise positioning
  • Multi-format output support (Markdown, HTML, Annotated Image)
  • Responsive design that scales perfectly across all devices

Quick Start

1. Download Model

Download the DeepSeek-OCR model from one of these sources:

• ModelScope: https://www.modelscope.cn/models/deepseek-ai/DeepSeek-OCR
• Hugging Face: https://huggingface.co/deepseek-ai/DeepSeek-OCR

2. Installation Options

Option Local Installation

1. Install Dependencies

   conda create -n deepseek-ocr python=3.12.9 -y
   conda activate deepseek-ocr
   pip install -r requirements.txt
   # Install Torch First
   pip install flash-attn==2.7.3 --no-build-isolation
   

2. Configure Model Path

   Edit start_ocr_webui.py line 26:

   # Change this line to your model path
   self.model_path = '/path/to/your/DeepSeek-OCR'

3. Run Application

   python start_ocr_webui.py

4. Access WebUI

   Open browser: http://localhost:7860

3. Usage

1. Upload one or more images
2. Enter OCR prompt (or use preset prompts)
3. Click "Recognize" button
4. View results in the Results/Summary tabs

Requirements

• Python 3.12+
• CUDA-compatible GPU (recommended)
• 16GB+ GPU memory 4080+
• PyTorch with CUDA support

Features

• Multi-image batch processing
• Multiple OCR prompt presets
• Bilingual interface (English/Chinese)
• Docker deployment support
• Real-time processing progress

Common Prompts

• General OCR: Free OCR.
• Markdown: <|grounding|>Convert the document to markdown.
• Table: <|grounding|>Extract all tables and convert to markdown format.