See all results >
Table of contents

Release Notes - DynamsoftCaptureVisionBundle

3.4.2001 (04/24/2026)

Fixed

  • Fixed an issue where coordinate conversion could behave abnormally after large images were compressed.
  • Fixed an issue where abnormal DCEFrame imageData could cause errors in the toCanvas() method.

improved

  • Optimized the behavior of the Dynamsoft Camera Enhancer torch and other camera-related features.

3.4.2000 (04/21/2026)

Highlights

AI-Powered Barcode Detection and Decoding

  • PDF417 Localization Model – Introduces the PDF417Localization neural network model for improved detection of PDF417 barcodes, especially under challenging conditions.

  • Code39/ITF Decoding Model – Adds the Code39ITFDecoder model for enhanced decoding of Code 39 and ITF barcodes under blurred or low-resolution conditions.

  • Deblur Models for 2D Barcodes – Adds the DataMatrixQRCodeDeblur and PDF417Deblur models to provide more effective recovery from motion and focus blur for DataMatrix, QR Code, and PDF417 barcodes.

ECI (Extended Channel Interpretation) Support

  • ECI Information Return – Adds support for retrieving Extended Channel Interpretation (ECI) data from barcodes. The new ECISegment interface, along with the eciSegments property in the BarcodeResultItem and DecodedBarcodeElement interfaces, enables access to character encoding information embedded in barcodes.

  • ECI-Based Text Interpretation – Adds support for interpreting ECI segments during barcode decoding, improving compatibility with international character sets.

Performance Improvements

  • On-Demand Model Loading – Implements lazy loading for AI models, reducing initialization time by loading models only when first needed.

  • Smart Model Selection – Models are now loaded based on configured barcode formats, minimizing memory usage by excluding unused models.

  • Improved Confidence Scoring – Enhances confidence score calculation for results from neural network models, providing more accurate quality indicators.

  • DPM Barcode Optimization – Improves recognition rate for Direct Part Marking (DPM) barcodes commonly used in industrial and manufacturing environments.

Identity Document Processing

  • Enhanced Passport Processing – Improves document edge detection accuracy for passport documents through optimized processing workflows.

  • Portrait Zone Detection – The MRZLocalization model now supports detecting portrait zone on identity documents, enabling automatic extraction of photo regions.

  • New DynamsoftIdentityUtility Module – Introduces a dedicated module for identity document processing, including the IdentityProcessor class with FindPortraitZone() method for precise portrait positioning from passports and ID cards.

New

Changed

  • Barcode text encoding fallback changed from UTF-8 to ISO-8859-1 when no ECI information is present in the barcode.

  • All methods under the ImageProcessor class now additionally accept DSImageData as an image input type.

  • All methods under the ImageIO, ImageProcessor, and ImageDrawer classes have been changed to static methods.

  • CodeParserModule.loadSpec() now returns a Promise<ErrorInfo> instead of void.

  • Updated default value of compensation parameter in convertToBinaryLocal() from 0 to 10.

  • convertToBinaryGlobal() and convertToBinaryLocal() of ImageProcessor class now support color, binary and grayscale images as input.

  • Parser resource files (.json) have been consolidated into encrypted .data files for improved security and simplified distribution:
    • AADHAAR.data, AAMVA_DL_ID.data, GS1_AI.data, MRTD.data, SOUTH_AFRICA_DL.data, VIN.data
  • CameraEnhancer.open() now returns Promise<void> when operating in singleFrameMode.

Removed

Fixed

  • Fixed incorrect coordinate in barcode result when using neural network models with a specified region.

  • Fixed crash and hang issues that could occur in certain scenarios.

  • Fixed various minor bugs and improved overall stability.

3.2.5000 (12/16/2025)

This release includes security maintenance updates to ensure continued protection of the product.

Security Updates

  • Updated third-party libraries to incorporate the latest security fixes.

3.2.4000 (11/11/2025)

Fixed

  • Fixed a bug that could cause a crash when using model localization with a 2D barcode.

3.2.2000 (11/04/2025)

🎉Milestone Release

Version 3.2.2000 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.

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.

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.

API Changes

New

  • Added the switchCapturingTemplate() method to the CaptureVisionRouter class to enable switching the active capturing template during the image processing workflow.
  • Added methods toBlob(), toImage(), and toCanvas() to the originalImageResultItem interface for flexible image data conversion.
  • Added the SetGlobalIntraOpNumThreads() method to the CaptureVisionRouter class to configure the global thread count for model inference.
  • Added the clearDLModelBuffers() method to the CaptureVisionRouter class to manually release memory occupied by loaded models.
  • Added callback functions onSpecLoadProgressChanged() and onWasmLoadProgressChanged() to monitor resource loading progress.
  • Added the convertToContainCoordinates() method to the CameraEnhancer class to convert coordinates from fit: cover to fit: contain mode.
  • Added the LM_NEURAL_NETWORK enumeration to EnumLocalizationMode.

Updated

  • Changed the default value of MaxThreadsInOneTask from 4 to 0.
  • Changed the default value of IncludeTrailingCheckDigit from 1 to 0.

Deprecated

  • Deprecated the DeblurModelNameArray argument in the DeblurModes parameter. Use ModelNameArray instead.
  • Deprecated the appendModelBuffer() method in the CaptureVisionRouter class. Use appendDLModelBuffer() instead.

3.0.6001(08/28/2025)

Fixed

  • Rebuilt the WASM file to resolve a performance degradation issue.
  • Added back missing type definitions for documentSettings and labelSettings.

3.0.6000(08/14/2025)

New

  • Base64 String Conversion: Enabled bidirectional conversion between ImageData and Base64 strings.

  • Template Version Validation: Introduced version checking for templates to prevent compatibility issues and mismatches.

Improved

  • White-on-White Document Recognition: Optimized accuracy for document detection in scenarios where white paper is placed on a white background.

Changed

  • License Validation Behavior: Instead of stopping execution immediately on an invalid license module, the library now continues processing and returns results from modules with valid licenses. An error is still reported to indicate the license issue.
  • StartCapturing now stops immediately and returns a license-related error code if none of the tasks have a valid license, instead of proceeding and returning an empty result.

Fixed

  • Fixed various minor bugs and improved overall stability.

3.0.3001(06/09/2025)

Fixed

  • Fixed an issue where the results from the last unprocessed frame were still rendered on the page after calling the captureVisionRouter.stopCapturing() method.

3.0.3000(06/05/2025)

Highlights

Workflow Improvements

  • Restructured the parameter control hierarchy at all levels for finer scope definition and more granular process management, with the stage level newly added.
  • Enabled custom combinations and sequences of sections, increasing flexibility and operational customization under specific conditions.
  • Redesigned document normalization sections to better accommodate diverse document processing operations.

Deep Learning Integration

  • Improved the reading rate of 1D barcode by introducing a new deblurring deep-learning model.
  • Enhanced text recognition capabilities with deep learning-based text-line recognition.

Algorithm Enhancements

  • Enabled deduplication at the Region of Interest (ROI) level to consolidate results from multiple tasks.
  • Enhanced the text recognition workflow by integrating improved multi-step recognition processes and validation methods.
  • 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.

Engineering Optimizations

  • Simplified the loading and configuration of WASM resources, reducing the overall size of the WASM files.
  • Added the dcvData package to host all potentially used resource files (such as machine learning models, code parser resources, preset template files, etc.).
  • Implemented conversion functionality between ImageData and image files, including both local and in-memory files.
  • Added utility functions for basic image processing tasks.

This page is compatible for: