Change Log🔗
Changes in Eocortex 4.5.102🔗
12/02/2025
Video analytics
Added support for NVIDIA GeForce RTX 50xx video cards by neural network modules of analytics.
The operation of the License Plate Recognition and Object Classification and Counting modules of video analytics has been optimized.
Now, the same hardware can process more cameras with enabled modules.
The video card load limit at which frame skipping begins in the module (accuracy degradation) has been increased from 60-70% to 90-100% (provided that the CPU is also not loaded to maximum capacity).
Improved operation of the License Plate Recognition module:
The method for recognizing vehicle brands has changed — it is now recognized by its logo and body shape (previously only by body shape). This approach significantly reduces the number of false vehicle brand recognitions.
Two new brands have been added: Daihatsu and Tesla.
Improved module performance when running on the CPU.
Improved accuracy of the Fall Detection, Emergency Vehicle Detection, Abandoned Objects Detection modules.
Improved accuracy of hard hat and safety vest detection in the Uniform Detection module.
Increased calculation accuracy in the Unique Visitor Counting module due to the following improvements:
The quality of detection for both horizontal rotation and various tilts of faces has been improved.
A parameter has been added to the report indicating whether rotated faces should be considered unique.
Fixed a bug that prevented automatic reports from being generated in the Object Classification and Counting module under certain circumstances.
Cameras and devices
Fixed the lack of receiving motion events from Uniview cameras.
The PTZ control for Panasonic WV-SPN-631 cameras with a connected external pan/tilt mechanism has been integrated.
Mobile applications
Fixed a bug that caused the Mobile iOS client Eocortex application to fail to launch when opened via a link from a browser.