Release notes

SentiVeillance 9.1 SDK

Subscribe for newsletters to get notifications about the releases.

SDK-related changes:
- Fixed a rare memory leak when using Vehicle-Human and License Plate Recognition algorithms.
Samples:
- Added SurveillanceImageServiceSampleCS sample which allows processing images by using REST API.
- Updated SurveillanceSampleCS to allow exporting each subject as single image.
- DetectFromImage tutorials now only obtains required license based on selected modality.

Algorithms’-related changes:
- Added template matching support for Mexico, Vietnam, Singapore, Malaysia and Indonesia license plates.
- Added recognition of Ä, Ö, Ü characters.
SDK-related changes:
- Added analytic triggers (Tripwires and Regions), allowing to generate events if subjects cross specified lines; enters, exits or dwells inside a region for specific time.
Samples:
- Updated SurveillanceSampleCS to support analytic triggers.
- Updated SurveillanceSampleCS to remember recently opened samples and selected modalities.

Algorithms’-related changes:
- Improved face quality estimation.
- Added license plate occlusion estimation.
- Added template matching support for New Zealand, Sri Lanka.

Algorithms’-related changes:
- Biometric algorithms from MegaMatcher 13.1 product line are now used.
SDK-related changes:
- Added support for processing using Intel GPUs/
Samples:
- Fixed an issue in Java sample with showing mirrored images.

Algorithms’-related changes:
- Improved License plate recognition accuracy.
- Added template matching support for Oman, Kuwait, Jordan, Bahrain, Qatar.
SDK-related changes:
- Allow to specify search areas when detecting from images with NSurveillanceEngine.
- Fix NCamera now correctly returns whether it supports returning NVideoSample.
Samples:
- Allow to search through priority countries in SurveillanceSampleCS.
- Allow to export all events in SurveillanceSampleCS.
- Allow to filter exported events by source in SurveillanceCS.
- SurveillanceSampleCS now exports more information to CSV file: vehicle tags, if license plate from watchlist and its owner.
- Tutorials now have option to enable GPU usage.
- Small fixes in surveillance-web-sample.

Algorithms’-related changes:
- Added support for Arabic license plates.
- Added template matching support for Egypt, Morocco, Saudi Arabia license plates.
SDK-related changes:
- Fix cases where best license plate has no assigned value.
- NSurveillanceEngine now throws more descriptive exception if data files are missing.
Samples:
- Added surveillance web sample using java and react framework.
- Added license plate watchlist to SurveillanceSampleCS.
- Added more event filters to SurveillanceSampleCS: in watchlist, not in watchlist, must have face, must have license plate.
- Improved faces watchlist user interface in SurveillanceSampleCS.
- Surveillance samples now has option to start all sources at startup if ‘–start’ command line argument is provided.
- Fixed small issues in java surveillance-sample.

Algorithms’-related changes:
- Tuned default face algorithm properties for better operation.
Samples:
- Removed thermal camera samples.
- Fixed export functionality in SurveillanceSampleCS.
- Fixed crash and functionality issues in Java surveillance-sample.

SDK-related changes:
- Fixed incorrectl check if the object is in specified search area when CheckSearchAreaByObjectCenter parameter is set.
- Fixed a bug with wrong object color returned when objects tracked for short duration.
Samples:
- Fixed SurveillanceSampleCS events export error.
- Fix surveillance-sample memory leaks and other minor issues.

Algorithms’-related changes:
- Face recognition algorithm updated to the one used in VeriLook/MegaMatcher 13.0 SDK.
- Improved object tracking algorithm.
- Improved ALPR algorithm.
- Vehicle model recognition algorithm updated with newer models and improved accuracy.
- Added feature to recognize person’s age group (Child, Teenager, Adult, Senior) when using vehicle-human modality.
- Added template matching support for Kyrgyzstan, Tajikistan, El Salvador license plates.
SDK-related changes:
- Model optimization can now return progress.
- Region of Interest is replaced with search areas, allowing for more precise configuration for selecting analyzed area.
- Added feature to return formatted license plate value.
- Changed frame decoding to only decode frames, which will be used in further processing, please update your applications if you are using ImageCaptured callback.
- Add support for NVIDIA Jetson devices (Requires JetPack 5.1.1).
Samples:
- Updated samples to include new features.
- Removed VB.NET samples and tutorials.