What's new in 24.09 for Analytics
Enhanced efficiency of the Analytics algorithms
Work of Analytics cases: boosting accuracy in precise detection and identification
By employing a novel feature-based analysis approach, we have successfully addressed the issue of false detections caused by complete camera occlusions and misidentification of static objects. Optimized algorithms enable faster, more accurate detection of moving objects within frames, allowing for precise object segmentation. By extracting unique features from cropped images, our system can confidently compare objects with previous frames.
Zone intrusion detection and Line crossing detection analytics cases are enhanced in the following way:
By employing an object-oriented approach, the system can now accurately and equally detect people, vehicles, and animals. This enhancement has significantly accelerated and refined event localization while completely eradicating false positives. Our ongoing efforts are dedicated to further refining the classification, search, and filtering processes for these objects.
Detected objects are now visually highlighted in the image captured during the analytics event, allowing the user to instantly identify the event's source. The image of the object is transmitted as a crop, simplifying the user's search for analytics events by image in the web interface under the Events tab > Analytics events.
The accuracy of line crossing detection has been improved for Visitor counting and Line crossing detection analytics cases. This was achieved by optimizing algorithms to eliminate false positives near the line.
We are pleased to announce advancements in the Visitor counting analytics case, that enhance the precision of age and gender identification. Previously, the system made a single prediction of age and gender for each frame in which a visitor was detected. Our innovative approach now involves averaging predictions across a visitor's entire journey. As a result, we have achieved a significant improvement in the accuracy of determining visitor age and gender and reduced the impact of random errors.
A new version of Model Server (the MVP stage) – a powerful analytics server that ensures high stability, performance, and scalability.
Zoning functionality is now introduced for Analytics workers. Administrative tools for this feature will be available in upcoming releases. This new solution allows for individual analytics workers to be tied to specific media servers based on zones. The need for this solution arises from the requirements of distributed branches (VMS clients) using their own infrastructure. Zoning enables the local installation of media servers and analytics workers (either individually or in groups), allowing for operation with a single VMS without being limited by insufficient bandwidth when interacting with other workers.