Why does the Wisenet WAVE AI Plugin freeze or drop streams?
Answered by ANPR AI·Sourced
Short answer
When utilizing the WAVE AI Plugin, adjusting the "AI Tracking Speed" too high exponentially increases the computational burden on the server. Setting the tracking speed to "Very Fast" for vehicles can severely reduce the number of AI streams the server can handle, leading to dropped frames and system freezing.
Explanation
The Wisenet WAVE AI Plugin allows operators to adjust the frame rate at which the artificial intelligence processes incoming video data. By default, this is set to 5 FPS, which is heavily optimized for standard surveillance and walking humans. However, when integrators attempt to tune the system to track fast-moving targets like cyclists or high-speed vehicles, they often increase the AI Tracking Speed to "Fast" (10 FPS) or "Very Fast" (15 FPS).
While this higher frame rate improves the accuracy of high-speed tracking, it essentially doubles or triples the AI processing load per camera. If a server was designed to handle 20 cameras at 5 FPS, switching them to 15 FPS will instantly exceed the server's compute budget. This results in severe performance degradation, dropped AI streams, or complete freezing of the WAVE client interface as the CPU struggles to clear the processing queue.
Things to check
- Review the AI Tracking Speed setting for all cameras utilizing the AI Plugin; ensure they are not unnecessarily set to "Very Fast".Monitor the host server's CPU and memory utilization in the Task Manager while the AI Plugin is actively processing.
What to do
Reduce the AI Tracking Speed back to the default 5 FPS for any cameras that do not strictly require high-speed vehicle tracking. To further optimize server performance, configure "Excluded Areas" within the plugin settings to mask out busy roads, swaying trees, or adjacent sidewalks; this prevents the AI from wasting processing power analyzing persistent, irrelevant movement in the background of the scene.
ANPR Edge
Skip the dedicated ANPR camera. Read plates from the cameras you already have.
ANPR Edge is plate-recognition software that runs on a PC at your site and works with ordinary IP cameras, so plate reads no longer depend on one vendor's camera, NVR or firmware.
- Works with standard IP cameras over RTSP
- One license per location covers every camera
- Runs on a Windows PC or Mac you own