Skip to content

Release notes - CVEDIA-RT - 2024.2.0

As usual, every new release brings updated models, new features and improvements, and bug fixes.

Please check below in detail what's new.

CVEDIA-RT Plugin 2.1 for NX Witness and white label variants

  • Split the Windows Installer into two separate installers. One for RT Studio and RT Server, and one for the RT Plugin for VMS.
  • The installer for CVEDIA-RT Plugin for VMS now supports Nx Meta, DW Spectrum, Hanwha WAVE, Piko VMS, and Alpha VS.
  • Added AlphaVS support for docker installs (--avs)
  • Hardened the Windows Installer to handle a previous manual installation of CVEDIA-RT Plugin for VMS.
  • Improved the Windows Installer to install missing dependencies.
  • Added command line options to run the installation of CVEDIA RT Plugin without user interaction on Windows: Installer CLI Options

CVEDIA-RT Plugin 1.0 for Milestone

First release of our plugin for Milestone. It offers the same analytics included in the plugin for NX Witness:

  • Instrusion Detection
  • Area enter/exit
  • Line crossing
  • Loitering
  • Crowding Detection

The events are searchable by event and class, and are available for the configuration of custom rules.

For more information please check the installation guide. The plugin is available on

SecuRT Solution 12.0

  • Improved object/left removed feature
  • Object Movement Estimation using NN is available now. To enable it, set Global.ObjectMovementRegressor.enabled to true.
  • All network URIs have been moved to a single place in base_config.json. This makes it easier to update them whenever needed.
  • The new default input size for the PVA detector is now 256. According to our validation tests, this model offers the same accuracy as the 320, but with better performance. It can be reverted back by setting Detector.preset_values.MosaicInference.Detector/model_file to pva_det_mosaic instead of pva_det_mosaic_256.

RT Studio

RT Studio Bug fixes

  • Fixed a bug that prevented opening the zone configuration settings when two or more zones were defined.
  • Fixed the cloning of a demo instance when no user instance have been already created.
  • Fixed concurrency issues when using RT Studio and REST API to manage instances. For example, an instance can now be started from the Web Panel and stopped from RT Studio.


  • Added NVIDIA video decoder support.
  • Added --trial option on Docker Linux-based installs to activate the trial license.
  • Added support for Memryx models
  • OpenVino models are now built against OpenVino 2024.1
  • TensorRT X86 models are now quantized
  • Added support for Hailo drivers 4.17
  • Updated documentation on license management
  • Added reference open-source unencrypted models in ONNX format: onnx://generic/rgb/medium/yolov6
  • Partially downloaded or corrupt models will be redownloaded when RT restarts.

JSON Event Schemas

  • Added a new schema for Fallen Person events
  • Removed the object class from the Armed Person event schema
  • Renamed event-armed-person-enter to event-armed-person
  • Renamed event-armed-person-exit to event-armed-person-end
  • Deleted old deprecated schema definitions
  • Renamed event-attribute-v1.json to attribute-v1.json (the schema id was already just attribute)
  • Renamed best-crop to crop event

You can access all the schema definitions at


  • Updated public CVEDIA-RT-SDK to reflect the latest changes on
  • CVEDIA-RT python bindings are now published as .whl packages on
  • Major refactoring of the C API and REST API focused on unifying and simplifying the overall structure. In particular, many functions and endpoints have been moved from SecuRT to RT Core.
  • Published REST API documentation: REST API.
  • Added ability to push compressed frames (JPG and PNG images) using REST
  • Added endpoint to define a mask in SecuRT REST API
  • Input and export endpoints have been moved from SecuRT API to Core RT API
  • The REST API endpoints and C API methods that return instances now include additional info: readOnly, autoStart, autoRestart, and systemInstance. systemInstance is true for the preinstalled demo that comes together with CVEDIA-RT.
C API changes New C API functions
The following functions are removed from SecuRT C API and moved into the Core C API
securt_get_instance_id -> core_get_instance_id
securt_get_instance_name -> core_get_instance_name
securt_get_version -> core_get_version
securt_set_detection_sensitivity -> core_set_detection_sensitivity
securt_set_movement_sensitivity -> core_set_movement_sensitivity
securt_get_track_count -> core_get_track_count
securt_get_frames_processed -> core_get_frames_processed
securt_get_latency -> core_get_latency
securt_get_current_frame_rate -> core_get_current_frame_rate
securt_set_frame_rate_limit -> core_set_frame_rate_limit
securt_get_frame_rate_limit -> core_get_frame_rate_limit
securt_get_recommended_frame_rate -> core_get_recommended_frame_rate

securt_set_blocking_readahead_queue -> core_set_blocking_readahead_queue
securt_set_config_value -> core_set_config_value
securt_set_statistics_mode -> core_set_statistics_mode
securt_set_metadata_mode -> core_set_metadata_mode
securt_set_render_preset -> core_set_render_preset
securt_set_auto_restart -> core_set_auto_restart
securt_set_detector_mode -> core_set_detector_mode
securt_set_debug_mode -> core_set_debug_mode
securt_set_tentative_tracks -> core_set_tentative_tracks
securt_set_sensor_modality -> core_set_sensor_modality
securt_set_diagnostics_mode -> core_set_diagnostics_mode
securt_set_input_to_rtsp -> core_set_input_to_rtsp
securt_set_input_to_manual -> core_set_input_to_manual

securt_create_context -> core_create_context
securt_delete_context -> core_delete_context
securt_reset -> core_reset
securt_is_alive -> core_is_alive
securt_need_data -> core_need_data
securt_is_instance_running -> core_is_instance_running
securt_enable_remote_mode -> core_enable_remote_mode
securt_disable_remote_mode -> core_disable_remote_mode
securt_find_remote_server -> core_find_remote_server
securt_start -> core_start
securt_stop -> core_stop
securt_restart -> core_restart
securt_free_string -> core_free_string

securt_push_frame -> core_push_raw_frame
securt_push_h264_frame -> core_push_h264_frame
securt_push_h265_frame -> core_push_h265_frame
securt_enable_hls_output -> core_enable_hls_output
securt_disable_hls_output -> core_disable_hls_output
securt_enable_rtsp_output -> core_enable_rtsp_output
securt_disable_rtsp_output -> core_disable_rtsp_output
securt_consume_events -> core_consume_events
The following functions undergone some changes while moving:
securt_update_instance_options -> core_update_instance_options -- with more options
securt_get_instance -> core_get_instance -- provides more information (sensor modality, detection sensitivity etc) now
The following functions are renamed
secure_create_license_plate_access_control_area -> securt_create_license_control_area
REST API changes InstanceRead and InstanceWrite models are now expanded with additional fields including detector sensitivity, sensor modality and so on. This affects the create instance endpoint. Endpoints removed from SecuRT REST API:
Endpoints added to Core REST API:
The following endpoints are removed from Securt REST API and moved to Core REST API
GET /v1/securt/instance/{instanceId} -> GET /v1/core/instance/{instanceId}
The following endpoints are added to Core REST API
PATCH /v1/core/instance/{instanceId}

CVEDIA-RT Bug fixes

  • Fixed missing rtconfig.json at persist folder on docker installs
  • Fixed possible dupe model delivery on platforms with CUDA compute 8.6+
  • Fixed issue with not all USB ports working on cvedia-box images
  • Fixed issue with not all ETH devices working on cvedia-box images
  • Fixed the capping of lines in the terminal