Changelog

Optical character recognition filter release notes

v0.5.0 - 2024-04-30

Enhanced documentation with feature descriptions:
- Added detailed topic-based processing capabilities
- Documented flexible output options (JSON file and metadata forwarding)
- Added environment variable configuration section
- Improved configuration reference table with all available options
Added example JSON outputs for both file and metadata formats
Added multi-camera system use case

Added support for topic_pattern configuration to selectively OCR frames based on topic name matching.
- New environment variable: FILTER_TOPIC_PATTERN
- Supports regular expressions to match topics.
Skips frames if their topic name does not match the specified pattern.

Updated documentation:
- Added detailed class and method docstrings.
- Clarified configuration attributes and processing steps.

OCR filter can now optionally forward detected texts into frame.data['meta']['ocr_texts']
- Controlled via the forward_ocr_texts setting (default: true)
- Allows downstream filters to consume OCR results without reading from external files

OCR filter writing to output JSON file is now fully optional
- Controlled via the write_output_file setting (default: true)
- When disabled, no file I/O is performed
OCR filter now supports overriding any of its configuration from environment variables
- Fields like ocr_engine, ocr_language, debug, forward_ocr_texts, write_output_file, etc. can be overridden using FILTER_* environment variables
- Supports loading from .env files as well
Improved robustness of OCR configuration parsing
- Correct handling of booleans, lists (like ocr_language), and validation of engine choice
- Clear error messages if invalid environment variables are provided
Fixed redundant double OCR calls for Tesseract
- Reduced unnecessary processing when extracting text lines
Improved shutdown behavior in OCR filter
- Proper logging depending on whether output file writing was enabled
Internal improvements