UMBO CV AI CCTV Light – Field Report
- 30 Dec, 2020
Reviewing Light A.I. Performance in the Field
Light helps video alarm monitoring centers scale by helping operators focus on what actually might be threats. This white paper dives into the results of a head to head test between Umbo Light and a well-known video analytics package. Over a span of over a month, we recorded and individually reviewed each alert sent by both systems.
The results find that within the recommended distance, Light can identify humans within its regions of interest as well as the competitor. Light customers get A.I. powered human detection on par or better with the rest of the industry.
But Light really shines when bad weather, foliage, or other adverse conditions come into play, showing a 99% reduction in false alerts. Even during chaotic storms that cause other video analytics packages to trigger all sorts of alerts, Light continues to send accurate human alerts.
Four cameras connected to the competitor analytics package had their RTSP streams piped into both A.I. products for a month, August 24th to September 24th. A neutral third party handled the alert monitoring and response. This third party is an established video alarm management company that handles over half a million video alerts each month. They have had over a year’s experience with both products.
The four cameras show a variety of outdoor locations. Each reflect their own challenges for video analytics and security camera alarm monitoring operators. The alarm configurations for both the benchmark video analytics product and Light were set to be as similar as possible. Light’s configuration is shown alongside the actual view.
Camera 4 shows a parkway leading up to a garage. The operator has instructed the system to alarm when it detects any person walking through the alleyway in the far distance. It should be noted that the alarming zone in the alleyway is placed beyond Light’s recommended distance.