Video Analytics Package DetectorPack 3.14.1 Released
AxxonSoft has released a new version of the video analytics package for Axxon One. DetectorPack 3.14.1 introduces new detectors, features, and built-in neural networks. Video analytics performance has been optimized for various hardware platforms.
New Features
Area Occupancy Detector
This detector, based on a segmentation neural network, identifies objects visually distinct from the background and calculates the occupancy rate — that is, the percentage of the designated area covered by detected objects. The neural network detects any distinguishable objects within the specified size range. The detector can report the current occupancy level and trigger alerts when a configured threshold is exceeded.
This tool is designed to monitor zones prone to clutter during various activities — such as corridors, driveways, construction and production sites, and warehouse areas. Timely detection of excessive occupancy reduces incident risks and helps maintain workplace safety and fire regulations.
Standard Neural Networks for Human Pose Detector
The Human Pose Detector now supports multiple built-in neural networks out of the box — no pre-training required, ready for use in most typical scenarios. The standard networks — Nano, Middle, and Large — differ in detection accuracy and computational load.
Intel Multi-GPU Support
For server configurations with multiple Intel GPUs, users can now assign specific detectors to individual graphics processors. Additionally, the new Intel Multi-GPU mode automatically distributes the workload across available GPUs.
Both features are supported by most of AxxonSoft’s detectors, including the Neural Tracker, Neural Counter, Human Pose Detector, PPE Detector, and Fire & Smoke Detectors.
Track Lifespan
A new parameter has been added for several detectors, indicating how long a detected object remains within a defined area. Presence duration is displayed in reports. In future updates, this information will also be visualized directly in the camera view for real-time monitoring.
Enhancements and Optimizations
Performance of the Neural Tracker, Human Pose Detector, and Barcode Detector has been improved when running on NVIDIA GPUs. The Neural Tracker now includes options to ignore static objects (track only moving ones) and to disable color detection — the latter feature reduces computational load.
New countries have been added for license plate recognition, along with the ability to specify vehicle movement direction. Vehicle make and model recognition accuracy has also been improved.
A configurable recognition threshold has been added to the Fight Detector to reduce false alarms and fine-tune sensitivity based on scene dynamics. Face Detectors VA and TV now include configurable parameters: face occlusion threshold and open eyes threshold — further minimizing false positives. In addition, strict filtering during face recognition helps ensure compliance with local requirements.
Installation and Documentation
Axxon DetectorPack 3.14.1 is compatible with Axxon One version 2.0 and later. For a full list of changes, see the release notes. Get the basic software package and add-ons from the download page.