v1.0.5

Release v1.0.5

🚀 New Features

• DICOM File Support:
  • The application now fully supports reading and writing DICOM files.
  • This includes enhanced handling for DICOM-specific file operations, such as saving images and burning text regions with tailored visual adjustments.

🛠 Refactoring and Enhancements

• New Utility Class: ImageUtils:
  • Introduced a new utility class, ImageUtils, to encapsulate common image processing functions.
  • Functions previously embedded in TesseractTextDetector, such as outlining text regions, adding watermarks, masking regions, and saving images with metadata, have been moved to this new class.
  • This abstraction facilitates reuse across multiple detectors and prepares the application for easier integration of new detection modules.

🔧 Improvements

• Improved modularity and maintainability by centralizing image-related tasks.
• Enhanced DICOM compatibility with proper handling of extensions (.dcm and .dicom).

v1.0.4

What's New in This Release?

New Features

1. Quarantine Functionality

   • Added logic to quarantine suspect text regions for review.

2. Dynamic Parameterization

   • Configurable bounding box sizes, quarantine paths, and other detector behaviors.

3. Batch Processing

   • Directory-based image processing for streamlined workflows.

4. Bounding Box Constraints

   • Introduced min/max size limits for bounding boxes in Tesseract.

5. Flexible Action Handling

   • Options for watermarks and metadata to align with user goals.

Improvements

6. Centralized Configuration

   • New DetectorConfig class for unified parameter management.

7. Enhanced CLI

   • Added support for directory paths and configurable parameters, plus a --help command for usage instructions.

8. Detailed Reports

   • Individual discrepancy reports generated per image for better traceability.

9. Improved Exports

   • Processed images and metadata are now saved for enhanced traceability.

New Tools and Compatibility

10. Ollama Flag

    • Optional flag for resource-intensive Ollama processing.

11. Burn Action

    • Semi-transparent red box overlay for detected text regions.

12. Extended File Support

    • Added compatibility for PNG and DICOM image formats.

Fixes and Usability Enhancements

13. Automatic Output Folder Creation

    • Ensures required output directories are created automatically.

14. Python Script Fixes

    • Addressed bugs in existing scripts for smoother operation.

v1.0.3

VisionGuard v1.0.3 Release Notes

This update brings key fixes and improvements to VisionGuard:

• OllamaTextDetector Fix: Resolved issues with the detector to ensure proper model responses are returned.
  • Resolved an issue where the OllamaTextDetector was not returning a usable result.
  • The detector now properly retrieves and processes responses from the model, enabling it to output meaningful results.
• DetectionResult Update: Added support for storing raw responses, enabling better handling of diverse outputs.
  • Introduced a new constructor for DetectionResult that supports storing raw responses (e.g., from Ollama).
• Enhanced DetectorValidator: Generates detailed reports comparing Tesseract and Ollama results, including matched and unmatched regions.
  • Added logic to extract and log meaningful text strings from Ollama's raw response for analysis.

Upgrade to take advantage of these fixes and improvements! 🚀

v1.0.2

VisionGuard Release Notes

What's New:

1. OllamaHelpers: Introduced helper utilities for streamlined interaction with the Ollama vision model, simplifying setup and usage.
2. Detector Comparison: Added the ability to compare results between Tesseract OCR and the Ollama vision model, with a detailed report highlighting discrepancies.
3. Prompt Tester: Added a tool to test prompts for the Ollama vision model. Results are included directly in the comments for easy reference.

This release enhances VisionGuard's modularity and validation capabilities, making it easier to integrate and evaluate multiple text detection models.

v1.0.1

1. Rename TextPHIDetector:

   • Rename it to TextDetector to make it more general and modular, as not all text is necessarily PHI.
   • The modularity would allow the interface to handle different use cases beyond PHI detection.

2. Command-Line Options for User Choice:

   • Allow users to specify how to handle detected text:
     • Outline: Draw bounding boxes.
     • Burn Out: Mask text areas (black/red overlay).
     • Move to Folder: Save the modified image to a specific directory for review.

v1.0.0

• Text detection, bounding boxes, and modified images are working.
• The SPI structure is in place for future extensibility.