dynamsoft_capture_vision_flutter 3.2.3000

dynamsoft_capture_vision_flutter: ^3.2.3000 copied to clipboard

🎉Milestone Release #

Version 3.2.3000 introduces a series of AI-driven improvements designed to enhance barcode and MRZ detection accuracy, processing speed, and configuration flexibility.

This release focuses on practical performance gains for production environments across retail, logistics, manufacturing, and identity verification workflows.

✨ Key Highlights #

AI-Powered Barcode Detection and Decoding

• New Localization Models – Introduces OneDLocalization and DataMatrixQRCodeLocalization neural network models for improved detection of blurred / low-resolution 1D codes, or partially damaged DataMatrix/QR codes.
• Specialized Decoders – Adds EAN13Decoder and Code128Decoder models optimized for long-distance and motion-blurred decoding scenarios.
• Redesigned Deblur Model – The OneDDeblur model now provides more effective recovery from motion and focus blur.
• Configurable Model Selection – The new ModelNameArray parameter supports flexible model loading and fine-grained control for specific barcode types.

Precision and Processing Control

• Enhanced Deblur Methods – DM_DEEP_ANALYSIS now includes sub-level control with OneDGeneral, TwoDGeneral, and EAN13Enhanced options.
• Barcode Count Expectation – The new ExpectedBarcodesCount parameter enables format-specific quantity control and early termination in fixed-count workflows.
• Improved Region Detection – The new RPM_GRAY_CONSISTENCY mode provides more precise region extraction based on grayscale uniformity and local consistency for document and label processing.

AI-Powered MRZ Detection

• Neural MRZ Localization – The new MRZLocalization model improves region detection accuracy and delivers up to 42.7% faster processing for MRZ-based document workflows.
• Configurable Localization Control – The new LocalizationModes parameter allows configuration for text line detection.

Smart Document Capture

• Clarity-Based Frame Selection – Automatically selects the sharpest and highest-quality frame in live capture workflows.
• Cross-Frame Verification – Updated verification algorithms enhance result reliability.

Performance Highlights #

Barcode Workflows

• Up to 26.5% higher read rates under blur conditions with as much as 44% faster processing.
• Reliable decoding of DataMatrix and QR codes with missing or damaged finder patterns.
• Extended operational range beyond 75 cm for long-distance barcode scanning.

Document Workflows

• Improved performance in live video capture environments.
• Consistent document quality through clarity-based frame evaluation.
• Faster MRZ processing for high-throughput identity verification

Developer Notes #

• Backward Compatibility – Fully compatible with existing integrations; no code-level changes required for upgrade.
• Configuration Flexibility – Expanded parameter set allows comprehensive model configuration for scenario-specific tuning.
• Production Stability – All new models validated in enterprise environments.

New #

• Added a new method, switchCapturingTemplate, which allows switching templates dynamically during the image processing workflow.
• Added a new method, clearDLModelBuffers, to release memory by clearing buffered deep learning models.
• Added a new method, setGlobalIntraOpNumThreads, to configure the global number of threads used for model execution.
• Added a cameraToggleButton and cameraToggleButtonVisible, to the CameraView, allowing users to switch between the front and back cameras.
• Added AztecDetails, DataMatrixDetails, OneDCodeDetails, QRCodeDetails, allowing users to access detailed information for each barcode.
• Added ImageIO, ImageProcessor and ImageDrawer, allowing users to load, process, and draw images for easier image manipulation and visualization.
• Added the rawValue property to the ParsedField class, allowing users to access the original value of a parsed field.

Changes #

• Refactored the getOriginalImage API by moving it from CaptureVisionRouter to IntermediateResultManager.
• ImageManager has been deprecated.

• Improved barcode decoding performance.

  • Improved the reading rate of 1D barcode by introducing a new deblurring deep-learning model.
  • Improved the CODE_128 and DataMatrix DeepAnalysis algorithms for better decoding accuracy and performance.
  • Added support for new barcode types: CODE_32, MATRIX_25, KIX, and TELEPEN.
  • Added GS1 Application Identifiers (AI) support for improved code parsing capabilities.

• Added MRZ scanning support.

• Added document scanning support.

• Upgraded APIs to provide seamless integration with the new features.

• ⚠️ Breaking Changes:

• Removed all v1.4.0 apis and added new apis.