Skip to content

Release notes - CVEDIA-RT - 2023.5.1

Note

This release is not yet available to the public. If you require access to these updates, please contact us. We will provide download links and a complimentary activation license key.

NX-Witness Integration

  • Updated and improved the documentation on how to setup, activate, and configure CVEDIA-RT Pluing for NX Witness
  • Improved the license activation workflow
  • Improved the FPS warning to be less aggressive
  • Exposed an option in the Expert settings to disable the FPS warning

SecuRT V6.0

  • Added Face detection as a class on all existing events. The solution keeps analyzing the head pose until the event triggers to capture the best face angle.
  • Added support for exporting schema-based events using MQTT
  • Added support for exporting the output with anonymized faces
  • Updated the solution to use the latest PVA detector in RGB and Thermal
  • Fixed an issue that prevented to switch between RGB and Thermal detectors at runtime

AI Models

  • Improved Crowd Estimation model to reduce false positives and increase recall on large crowds
  • Improved Person-Vehicle-Animal (PVA) RGB and Thermal detectors

CVEDIA-RT Box

  • Added technical documentation for the management and configuration of the CVEDIA-RT Boxes

Engine

General improvements

  • Improved stability over long-run execution
  • Improved performances
  • Improved motion detection
  • Added support to download video files on demand, as input of an Instance
  • All the demo video files included in the release are now moved to the cloud
  • Implemented track reset when changing video in a playlist
  • Added support for and list of Input's URI in instance JSON configuration
  • Added support for logs rotation that can be configured in the global.config: 
    "max_log_file_size": 52428800,
"max_log_files": 3,

RTSP and GStreamer

  • Improved RTSP connectivity on various supported protocols and formats
  • Added a GStreamer tutorial on how to debug and address RTSP connectivity issues

Data Export

  • Added armed person detected event schema
  • Added support to export the output stream in HLS
  • Added support for the pre-processing of schema-based events using an external application or script
  • Added support for configuring the aspect ratio and the "zooming out" factor of the best-crop of the track: 
    "thumbnail_scale_back": 1.5,
"thumbnail_aspect_ratio": 1.0

Bug fixes

  • Fixed an issue running Yolov6 models on Jetson Nano + jetpack 4.6.x
  • Fixed "Invalid track/frame width/height received" when the input stream changes resolution
  • Fixed WriteData plugin unable to writeImage to uncreated folder