Skip to content

Release Notes - CVEDIA-RT - 2024.2.5

Plugin for NX Witness

  • Introduced new issue reporting methods (requires NX Witness 6) for False Positives and False Negatives via object thumbnails or custom bookmarks, with the ability to automatically send a video and open a support ticket. The reports will also contain the current plugin configuration, removing the need to create multiple screenshots.
  • Implemented a new type of excluding area that prevents new tracks from being created within that area, but allows other tracks to cross it without being lost as would happen with a masking area that blackens the input image.
  • Removed the camera rotation option, automatically aligning rotation settings with NX Witness camera configurations.
  • Added a status panel in the Plugin page for the NX Plugin. We now show the status related to the active streams, available streams and licensing.
  • Person thumbnails include faces as an overlay.
  • Enhanced clarity in the Statistics panel for inactive instances.
  • Improved plugin responsiveness, especially when stopping an instance.
  • Added support for Nx Go on Windows and Linux
  • Improved installer functionality:
    • Enhanced handling of NX Media Server.
    • Smoother updates to newer versions.
  • Fixed the issue of the notification alerting for the instance continuously restarting upon re-assigning a license.
  • Fixed an issue where the coordinates for the analytics were incorrect when the cameara was rotated in the NX Camera settings.

Plugin for Milestone

  • Enhanced performance and stability.
  • Introduced live-streaming during line and polygon editing.
  • Introduced Object Guarding, Tailgating, and Object Left Behind analytics.
  • Fixed and issue that prevented Object detector for not being properly updated when switching between Standard and Hi-Res performance modes
  • Improved installer functionality:
    • Enhanced handling of the Milestone Management Server.
    • Smoother updates to newer versions.

WebPanel

Web Panel General Settings

  • Introduced new analytics options for:
    • Tailgating
    • Object Guarding
    • Object Left
    • Line Counting
    • Area Occupancy
    • Crowd Estimation
    • Armed Person Area
    • Fallen Person Area
  • Inline help added for better user understanding of options and analytics.
  • Added a section to configure many instance properties such as:
    • Auto start
    • Auto restart
    • Profile (Standard / Hi-Res)
    • Camera orientation
    • Options to export appearance search and re-id features
    • Max AI resolution
    • AI tracking speed
    • Detection mode (Smart detection / Active detection)
    • Detection sensitivity
    • Movement sensitivity
  • Fixed bugs:
    • Username and password cannot be entered in the MQTT export settings
    • A host name cannot be entered in the MQTT export settings. It accepts only an IP.

SecuRT

  • Introduced new analytics and counting event event-count-changed:
    • Crowd Estimation Area
    • Occupancy Area
    • Counting Line
  • Introduced the Dwelling Area analytics that emits the event-dwelling event.
  • Added event-status-changed providing information about object status change: movement started, stopped, restarted.
  • Implemented functionality to export re-identification (ReID) features for people, faces, and vehicles as attributes.
  • License plates can now be exported as attributes of vehicle tracks.

  • Introduced a new exclusion area type that:

    • Prevents new tracks from being created within the specified area.
    • Allows existing tracks to cross the area without being lost.

    API Changes

    - The `exclusion_areas` endpoint used for masking areas has been renamed: `POST /v1/securt/instance/{instanceId}/masking_areas`
- The old endpoint will now manage the new exclusion areas:
    - `GET /v1/securt/instance/{instanceId}/exclusion_areas`
    - `POST /v1/securt/instance/{instanceId}/exclusion_areas`
    - `DELETE /v1/securt/instance/{instanceId}/exclusion_areas`

  • securt_test.py: Wrapped parameterized tests using dicts as enums to enable nicer test name printing

  • Fixed bugs:
    • Crowding Areas created using the SDK, REST API, and Web Panel defaulting to Loitering Areas.
    • Appearance search no longer includes color attributes for bicycles/motorbikes.

Engine

  • Default AI frame rate for new instances reduced from 10 to 5 FPS.
  • Inference Dispatcher is enabled by default on Tensor RT (NVIDIA GPUs) leading to better performance on systems with a large number of running instances.

  • Extended and Updated C API

    C API changes

    Added the following Core C API function:
- core_get_input_orientation

Added support for the following Core and SecuRT C API functions:
- core_set_input_orientation
- core_get_input_orientation
- core_set_input_pixel_limit
- core_get_input_pixel_limit
- core_is_instance_running
- core_set_blocking_readahead_queue
- core_get_config_value
- core_get_preferences
- core_get_decoders

- securt_set_performance_profile
- securt_enable_lpr
- securt_enable_face_detection

The signatures of the following Core C API functions have been modified to include a context handle, as part of the added remote mode support:
- core_get_decoders
- core_get_preferences

New unified core_push_raw_frame and core_push_encoded_frame.

The signature of Core C API function core_get_instance has been modified to allow the retrieval of two more instance fields:
- inputPixelLimit
- inputOrientation

  • Migrated mf.cvedia.com and kg.cvedia.com to central domain api.cvedia.com

  • Decoders can now output YUV directly
  • RT pipeline now supports YUV
  • Buffer class now support NV12/YV12 in addition to BGR
  • Added support for Jetpack 6.0 and Jetpack 4.6.2+
  • Linux builds now support ONNX inference backend.
  • Fixed bug:
    • Detector mode being overridden with Smart Detection upon loading an instance.
    • VCA plugin losing the license on update / reinstall / uninstall
    • Enabling autostart on a running instance, causes the log to be spammed with errors.
    • Frame rate limit on instances with RTSP input not being applied.
    • Using an array of videos as Input/URI in SecuRT is failing.
    • rtcmd inference evaluate command having issue when dealing with datasets where not all classes were present
    • Object track was collapsing to one of the corners of the view
    • rtcmd --version is not returning the proper version

RT Studio

  • Renamed the event "Person Intrusion" to "Armed Person Intrusion" when the gun attribute is enabled.
  • Fixed a bug causing base config entries to be erroneously saved in instance config files.

REST API

  • New Core REST API endpoints for configuration and diagnostics.

  • Enhanced retrieval of exclusion areas and analytics entities, now including:

    • Armed Person Area
    • Object Removed Area
    • Object Left Area
    • Fallen Person Area
    • Tailgating Line
    • Occlusion Area
    • Line Counting
    • Crowd Estimation

    REST API Changes

    The following Core REST API endpoints have been added:
- GET /v1/core/system/decoders
- GET /v1/core/system/preferences

The Core REST API endpoint PATCH /v1/core/instance/{instanceId} now allows for the modification of 3 more fields:
- blockingReadaheadQueue
- inputOrientation
- inputPixelLimit

The following SecuRT REST API endpoints have been added:
- POST /v1/securt/instance/{instanceId}/performance_profile
- POST /v1/securt/instance/{instanceId}/face_detection
- POST /v1/securt/instance/{instanceId}/lpr

  • Fixed bugs:

    • Incorrect REST API response for system configuration retrieval.
    • Face analytic not saving properly when created or modified via the REST API.
    • Core consume_events REST endpoint failing on auto-started instances
    • The tailgating configuration returned by a few endpoints doesn't return the crossing direction.

Known Issues

  • Milestone Plugin does not operate in remote mode. A fix will be provided in a future release.