PaddleOCR-VL Web UI

A modern, clean web interface for PaddleOCR-VL model built with Gradio.

Features

• 📑 Document Parsing: Full-page document analysis with layout detection (supports images & multi-page PDFs)
• 🎯 Element Recognition: Targeted recognition for text, formulas, tables, charts, and seals
• 🔍 Spotting: Detect and locate specific elements in images
• 🖼️ Base64 Image Support: Handles both URL and Base64-encoded images from API
• 📄 PDF Support: Upload or drop PDF files in Document Parsing mode
• 📱 Responsive Design: Modern UI that works on desktop and mobile

Installation

# Install dependencies
pip install -e .

# Or install directly
pip install gradio pillow requests pydantic-settings

Usage

# Run with default settings
python -m paddleocr_ui

# Or use the entry point
paddleocr-ui

# Custom API endpoint and authentication
python -m paddleocr_ui --api-url http://your-api-endpoint/predict --api-key your-key

# Custom host and port
python -m paddleocr_ui --host 0.0.0.0 --port 8080

# Create a public shareable link
python -m paddleocr_ui --share

# Enable debug mode to log API requests and responses
python -m paddleocr_ui --debug

# Show all available options
python -m paddleocr_ui --help

Configuration

Configuration is provided via command line arguments.

Command Line Options

Option	Default	Description
--api-url	http://localhost/layout-parsing	API endpoint URL
--api-key	(none)	API key for authentication
--host	127.0.0.1	Host to bind to
--port	7860	Port to bind to
--share	(disabled)	Create a public shareable link
--debug	(disabled)	Enable debug mode

Debug Mode

When debug mode is enabled (--debug), the application will log:

• Full request payload to stdout (with base64 image data truncated for readability)
• Full API response to stdout (with base64 image data truncated)

This is useful for debugging API issues and comparing requests with direct API calls.

API Response Format

The application expects API responses with the following structure:

{
  "errorCode": 0,
  "result": {
    "layoutParsingResults": [
      {
        "markdown": {
          "text": "# Document content...",
          "images": {
            "image1": "base64-encoded-string-or-url"
          }
        },
        "outputImages": {
          "visualization": "base64-encoded-string-or-url"
        }
      }
    ]
  }
}

Images in the response can be either:

• Base64 encoded strings: Will be converted to data URLs
• HTTP URLs: Will be used directly

License

MIT License