The neural detector detects and classifies various postures to track a typical behavior (unusual posture at the ATM, falling, raised arms during an attack, etc.). TRASSIR Pose Detector consists of a neural network, tracking, and scripts (you can add new poses to the classification). The neural network detects a person and finds the position of their body. By tracking the detected pose, the Pose Detector determines what the pose corresponds to at the moment (the average body position over 2 seconds), and classifies it using a script.
You can customize the size of detected objects, zones where pose identification is needed, sensitivity, quality and recognition speed:
Quality: the detector operates in two modes - simple and advanced; in advanced mode it recognizes more subtle poses using the high definition network and requires more powerful equipment.
Recognition speed: 6 fps, 12 fps, ongoing - in each frame; the fewer frames per second, the more difficult it is for the detector to recognize the pose.
Sensitivity: the higher this value, the more susceptible the detector is to changes in posture; a decrease in the indicator eliminates false positives.