VisionGuard

VisionGuard is a robust image processing and text detection system designed to process images, extract text, and apply various actions based on user-defined configurations. Leveraging the power of Tesseract OCR and external API integrations, VisionGuard provides a flexible and extensible platform for handling text detection tasks.

Features

• Text Detection:

  • Detect text regions and extract textual content using Tesseract OCR.
  • External API support via OllamaTextDetector for alternative text extraction methods.

• Configurable Actions:

  • OUTLINE: Highlight detected text regions with bounding boxes.
  • MASK: Obscure detected text regions for privacy purposes.
  • BURN: Apply a semi-transparent red overlay to detected text regions.
  • EXPORT_TO_FOLDER: Save processed images and metadata to a designated folder.
  • FLAG_FOR_REVIEW: Mark images as needing further review by adding a "QUARANTINE" watermark.

• Flexible Configuration:

  • CLI arguments for dynamic bounding box constraints and folder paths.
  • Supports customizable detectors and actions.

• Discrepancy Reporting:

  • Generate reports comparing results from multiple detectors.

Installation

Prerequisites

1. Java Development Kit (JDK): Ensure you have JDK 8 or higher installed.

2. Tesseract OCR:

   • Install Tesseract OCR on your system. For example:

     brew install tesseract
     # OR
     sudo apt-get install tesseract-ocr

   • Specify the Tesseract data path in CLI arguments or code configurations.

3. Dependencies:

   • Add the required dependencies to your pom.xml or equivalent build configuration:
     • Tesseract4J
     • Google Gson

Usage

Running the Application

1. Compile and run the project:

   javac -cp . com/tdiprima/visionguard/VisionGuard.java
   java com.tdiprima.visionguard.VisionGuard <directoryPath> <action> <outputPath> <reportPath> [options]

2. Example command:

   java VisionGuard input/ OUTLINE output/ report/ --minWidth=20 --minHeight=20 --maxWidth=500 --maxHeight=500

   # Displays the actions and parameters, below
   java VisionGuard --help

Actions

Action	Description
OUTLINE	Draw bounding boxes around detected text regions.
MASK	Obscure detected text regions.
BURN	Apply a semi-transparent red overlay to detected text regions.
EXPORT_TO_FOLDER	Save the processed image along with metadata to the specified folder.
FLAG_FOR_REVIEW	Add a "QUARANTINE" watermark and move the image to the quarantine folder.

Optional Parameters

Parameter	Description	Default Value
--ollama=true/false	Enable or disable OllamaTextDetector	false
--minWidth=X	Minimum width of bounding boxes.	10
--minHeight=Y	Minimum height of bounding boxes.	10
--maxWidth=A	Maximum width of bounding boxes.	500
--maxHeight=B	Maximum height of bounding boxes.	500

Project Structure

com/tdiprima/visionguard/
├── VisionGuard.java             # Main application entry point
├── DetectorConfig.java          # Configuration loader for CLI arguments
├── TextDetector.java            # Interface defining detector methods
├── TesseractTextDetector.java   # Tesseract OCR-based implementation
├── OllamaTextDetector.java      # External API-based implementation
├── DetectorValidator.java       # Utility for validating and comparing detection results
└── resources/                   # Resource files (e.g., Tesseract training data)

How to Contribute

Please see CONTRIBUTING.md

License

This project is licensed under the MIT License. See the LICENSE file for more details.