Skip to content

README_classifier.md

Latest commit

 

History

History
192 lines (149 loc) · 6.42 KB

README_classifier.md

File metadata and controls

192 lines (149 loc) · 6.42 KB

OCR Image Classifier

A powerful web-based tool for automatically classifying images using Optical Character Recognition (OCR) with Tesseract.js. This tool extracts text from images and automatically categorizes them based on configurable keyword rules.

Features

🖼️ Image Processing

  • Multi-format Support: Upload images in various formats (PNG, JPG, JPEG, etc.)
  • Batch Processing: Process multiple images simultaneously
  • High-Quality OCR: Powered by Tesseract.js for accurate text extraction

🌍 Multi-Language OCR Support

  • English + Traditional Chinese (default)
  • English only
  • Traditional Chinese only
  • Japanese
  • Korean
  • French
  • German
  • Spanish

🏷️ Smart Classification

  • Custom Rules: Create classification rules with keywords and colors
  • Automatic Categorization: Images are automatically classified based on extracted text
  • Visual Indicators: Color-coded classifications for easy identification
  • Real-time Updates: Classification updates as you modify rules

🔍 Interactive Features

  • Word Highlighting: Click on extracted words to highlight them on the image
  • Zoom & Pan: Navigate through images with mouse wheel zoom and drag pan
  • Bounding Boxes: Visual representation of detected text regions
  • Accordion View: Expandable list of all extracted words

📊 Export & Analysis

  • CSV Export: Download classification results in CSV format
  • Batch Results: Process and export results for multiple images
  • Status Tracking: Monitor processing status for each image

How to Use

1. Upload Images

  • Click "Upload Images" to select one or more image files
  • Images will be automatically queued for OCR processing
  • Processing status is displayed on each thumbnail

2. Configure Classification Rules

  • Add New Rule: Click "Add New Rule" to create a new classification
  • Keywords: Enter comma-separated keywords that trigger classification
  • Colors: Choose a color for each classification category
  • Priority: Rules are processed in order (first match wins)

3. Monitor Processing

  • Watch real-time progress as images are processed
  • Status indicators show: "queued", "processing...", "classifying...", "done", or "error"
  • Classified images display their category and color

4. Interact with Results

  • Select Images: Click thumbnails to view full-size images
  • Explore Text: Use the accordion to view all extracted words
  • Highlight Words: Click on words to highlight their bounding boxes
  • Navigate: Use mouse wheel to zoom, drag to pan

5. Export Results

  • Click "Download CSV" to export all classifications
  • CSV includes: Image Name, Classification, Extracted Text

Classification Rules

Default Rules

The tool comes with two example rules:

  • Invoice: Keywords: invoice,bill,receipt (Red)
  • Contract: Keywords: agreement,contract,terms (Blue)

Creating Custom Rules

  1. Click "Add New Rule"
  2. Enter a descriptive name (e.g., "Receipt", "Form", "Letter")
  3. Add comma-separated keywords that appear in that document type
  4. Choose a distinctive color
  5. Rules are automatically applied to all images

Keyword Tips

  • Use specific, unique terms for better accuracy
  • Include variations and synonyms
  • Avoid overly common words
  • Test with sample images to refine rules

Technical Details

OCR Engine

  • Tesseract.js: WebAssembly port of Google's Tesseract OCR engine
  • Worker-based: Non-blocking OCR processing
  • Language Models: Pre-trained models for multiple languages

Performance

  • Parallel Processing: Multiple images processed simultaneously
  • Memory Efficient: Images processed in chunks
  • Progress Tracking: Real-time status updates

Browser Compatibility

  • Modern browsers with WebAssembly support
  • Chrome, Firefox, Safari, Edge (latest versions)
  • Mobile browsers may have performance limitations

Use Cases

Document Management

  • Automatically categorize invoices, receipts, contracts
  • Organize scanned documents by type
  • Batch process large document collections

Data Extraction

  • Extract text from forms and surveys
  • Process handwritten notes and documents
  • Analyze image-based content

Quality Assurance

  • Verify document types match expected content
  • Identify misclassified documents
  • Monitor document processing workflows

Tips for Best Results

Image Quality

  • Use high-resolution images (300+ DPI)
  • Ensure good contrast between text and background
  • Avoid blurry or skewed images
  • Use consistent lighting

Keyword Selection

  • Start with specific, unique terms
  • Test rules with sample images
  • Refine keywords based on results
  • Consider document context

Language Selection

  • Choose the primary language of your documents
  • Use multi-language option for mixed content
  • Consider regional variations in text

Troubleshooting

Common Issues

OCR Not Working

  • Check browser console for errors
  • Ensure images are valid and not corrupted
  • Try refreshing the page and re-uploading

Poor Classification Accuracy

  • Review and refine keyword rules
  • Check image quality and text clarity
  • Verify language selection matches document language

Slow Processing

  • Reduce image resolution if possible
  • Process fewer images simultaneously
  • Check browser performance settings

Performance Optimization

  • Close other browser tabs
  • Use modern, fast browsers
  • Ensure stable internet connection
  • Process images in smaller batches

Browser Requirements

  • WebAssembly Support: Required for Tesseract.js
  • Modern JavaScript: ES6+ features
  • Memory: Sufficient RAM for image processing
  • Storage: Local storage for user preferences

Privacy & Security

  • Local Processing: All OCR processing happens in your browser
  • No Upload: Images are not sent to external servers
  • Temporary Storage: Images are stored only during the session
  • Data Export: Only processed results are exported

Future Enhancements

  • Machine Learning: Improved classification accuracy
  • Template Matching: Visual pattern recognition
  • Cloud Processing: Option for server-side processing
  • API Integration: Connect with external document systems
  • Advanced Export: PDF reports and detailed analytics

Version: 1.0
Last Updated: 2025
License: Open Source
Contributors: PDF Toolbox Team