Release Note 23.3
Download PDF: Release Note v.23.3.pdf
Included Plugins
- Age & Gender Detection
- Face Recognition
- Gun Detection
- Hard Hat Detection
- Heat Map
- LPR
- Motion Detection
- Object Detection
- Pose Estimation
- Smart Parking
- Smart Tracking System
- Smoke & Fire Detection
- Traffic Analytics
- UIC & Container
General Changes
- Improved analytics accuracy for Gun Detection, Hard Hat Detection, and LPR plugins.
- Added the ability to generate PDF and CSV reports for LPR and Face Recognition plugins.
- Enabled analytics to work based on keyframes to significantly reduce the decoding chip load. This feature is implemented for Heat Map, Motion Detection, Object Detection (with Object Counter trigger, provided Dwell Time is set to 0), Smart Tracking System, and Smoke & Fire Detection plugins.
- Unified CPU and GPU installation, with priority given to GPU. If no suitable GPU is available, the system will use CPU. Analytics can now be run on both GPU and CPU.
- For Linux Ubuntu 20.04, we have seamlessly integrated all analytics components into the Process Runner installer. Now all analytics components are automatically installed alongside the Process Runner, streamlining the Analytics System installation process.
- Intel CPU optimization.
- Improved system stability.
LPR / Smart Parking
- We have expanded the list of supported countries for license plate recognition, now including Argentina, Bolivia, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Japan, Nicaragua, Panama, Paraguay, Peru, Singapore, Thailand, and Uruguay.
- The accuracy of license plate recognition has been improved in Europe + CIS, Middle East, and USA + Canada + Mexico analytical models.
- Advanced functionality has been added for selecting analytics models and configuring more precise detection based on country-specific license plate templates. This feature is accessible in the "Advanced Settings" tab of the analytics settings.
- The system now supports recognition by zones, with the recognition zone information redirected via API in the metadata package.
- Motion-based analytics can now be configured in the "Hardware" tab of the analytics settings. Frames will only be processed by the neural network when motion is detected.
Age & Gender Detection
We've implemented the ability to manually set the interval between analytics data updates in the "Advanced Settings" tab.
Face Recognition
The Liveness feature has been reintroduced to check for artificial elements like mannequins, masks, and dolls in the frame.
Pose Estimation
We revamped analytics to accurately identify a person's posture in the frame by the following rules:
- Fallen Down
- Sitting
Object Detection
A new Crowd trigger has been added to send notifications about crowds of people within a specified area over a set duration, compatible only with the Person object type. The trigger considers the distance between objects to identify people gatherings in the frame.
UIC & Container
The ability to determine the detection zone has been added to the analytics settings.
Crowd Detection
Support for the Crowd Detection plugin has been discontinued, incorporating its functionality into the Object Detection plugin (Crowd trigger).
Smoking Detection
Support for the Smoking Detection plugin has been discontinued.