Skip to main content

Changelog

TextScan release notes

v0.1.7 - 2025-09-27​

Changed​

  • Updated documentation.

v0.1.6 - 2025-09-22​

Added​

  • Multi-topic processing - Process multiple video regions simultaneously using pub/sub topics
  • Data forwarding - New forward_upstream_data configuration to forward non-image frames
  • Main-first ordering - Ensures consistent output structure with 'main' topic first
  • Enhanced configuration normalization - Robust string-to-type coercion and validation
  • Comprehensive test suite - Added test_smoke_simple.py and test_integration_config_normalization.py
  • Environment variable support - FILTER_FORWARD_UPSTREAM_DATA for data forwarding control

Changed​

  • Improved process() method - Complete rewrite to handle multi-topic processing
  • Enhanced OCREngine.from_str() - Better handling of various input formats
  • Updated normalize_config() - Comprehensive configuration validation and type coercion
  • Updated documentation - README.md and docs/overview.md reflect new multi-topic capabilities
  • Updated usage script - scripts/filter_usage.py now demonstrates multi-topic processing

Technical Improvements​

  • Performance optimization - Efficient processing of multiple topics in single pass
  • Error handling - Robust validation and graceful error handling
  • Code consistency - Follows established patterns from other Plainsight filters

v0.1.5 - 2025-08-12​

Modified​

  • Lock Setuptools-SCM version

v0.1.4 - 2025-08-06​

Modified​

  • Support for model info
  • Updated dependencies

v0.1.3 - 2025-07-30​

Added​

  • Support for model info

v0.1.2 - 2025-07-14​

Changed​

  • Removed unnecessary imports
  • Updated openfilter version in pyproject

Added​

  • Feature that computes and sends ocr confidence score for each frame to metadata
  • Each entry in output jsons now include detected ocr_confidence
  • Support for ocr_language=eng for tesseract
  • Tests for ocr confidence score

v0.1.1 - 2025-05-22​

Changed​

  • Update dependencies

v0.1.0 - 2025-05-22​

Added​

  • Initial release of TextScan to extract text from image frames using OCR engines.
  • Dual OCR Engine Support:
    • Supports both tesseract and easyocr
    • Selectable via the ocr_engine config field
  • Multi-language OCR:
    • Accepts a list of language codes (e.g. ['en', 'zh']) via ocr_language
  • Output to JSON:
    • Writes OCR results per frame to a configurable output_json_path
    • Each entry includes frame_id and detected texts
  • Debug Mode:
    • Optional debug mode enables verbose logging
  • Frame-level Skipping:
    • Skips OCR for frames tagged with metadata flag skip_ocr: true
  • Tesseract Binary Configurable:
    • Customizable Tesseract binary path via tesseract_cmd (defaults to packaged AppImage)
  • Safe Streaming Output:
    • Results are flushed to disk in real-time to avoid data loss (future enhancement planned for configurable flush behavior)
  • Support for topic_pattern configuration to selectively OCR frames based on topic name:
    • New env var: FILTER_TOPIC_PATTERN
    • Supports regular expressions to match topics
  • Optional forwarding of OCR results into frame.data['meta']['ocr_texts'] via forward_ocr_texts (default: true)
  • Optional file output via write_output_file (default: true)
  • Support for .env and FILTER_* environment variable overrides
  • docs/~internal directory added (excluded from production)
  • Enhanced documentation:
    • Detailed topic-based processing and environment variable config
    • Updated examples and configuration table
    • Multi-camera use case

Changed​

  • Updated default config values to match implementation
  • Improved configuration parsing:
    • Proper handling of booleans, lists, and validation
    • Clear error messages for invalid env vars
  • Improved shutdown behavior and logging
  • Reduced redundant OCR calls in Tesseract
  • Improved FPS performance through internal tweaks
  • Improved documentation readability and structure

Fixed​

  • FILTER_OCR_LANGUAGE is now parsed correctly
  • Fixed incorrect detection of PROMO labels as line items
  • Fixed logic for transaction_status updates (ENG-1373)
  • Replaced external_id_type with external_id (ENG-1371)
  • Fixed double OCR processing and shutdown behavior issues
  • Fixed template for temporary JSON Formatter

Feature​

  • Added frame_skip variable to improve processing efficiency

Internal​

  • Internal improvements and refactoring

Experimental​

  • Temporary JSON Formatter added:
    • Enabled via transform_to_mqtt flag
    • Will be removed after integration with JSONOutputFormaterFilter

Payload Enhancements​

  • video_chunks added to the outgoing payload