False Alarm Filtering
By utilizing proprietary computer vision algorithms, Scylla AI video analytics triggers alerts only in case of validated threats, impressively minimizing false positives by up to 99.95%Book a demo
An alert is classified as a true alarm when the prediction of AI corresponds to reality (i.e. the object of interest is correctly identified, the action sought after is detected, etc.). A false positive is a case when the alert is triggered by mistake. Due to the essentially probabilistic nature of AI, the latter are inevitable in most cases. However, due to the elaborate AI and machine learning behind the False Alarm Filtering System, Scylla FAF can meet any level of production-grade industrial standards. Moreover, we are continuously improving Scylla AI modules where they are retrained on mistakes to make sure the number of false alarms goes even further down with time to achieve this impressive 99.5% benchmark.
It is not difficult to assemble from off-the-shelf deep learning modules to get a result of around 90% accuracy. The problem arises when we need more of it. Let’s consider an example: a client, e.g. a monitoring station, receives 50,000 events per time unit from their video surveillance network. If we assume 90% filtering efficiency, the rest remains to be handled, that is 10%, specifically 5,000 events. The operator has to qualify and process each of these 5,000 events. On the other hand, even at 99+% efficiency, which is provided by the Scylla FAF module, with the same assumptions, there is still 1%, or 500 events to handle. And it is this difference that is significant for effective operation and cost-saving.
It is possible to define settings for each camera to improve the relevance of alerts back to the dashboard.
Due to the statistical nature of any automated solution, a few False Alarms inevitably may happen. However, by utilizing proprietary computer vision algorithms Scylla's powerful AI video analytics constantly learns from mistakes and the number of False Alarms will steadily and gradually decline. Indeed, we continuously improve Scylla AI modules to make sure the number of false alarms goes even further down with time to achieve this impressive 99+% accuracy. Also consider that in any production setting Scylla AI analytics proved to be more accurate than the average human-factor errors typically accompanying routine surveillance tasks. Think about it, filtering 99.9% of alarms that would otherwise end up on the operator's dashboard means that instead of a meticulous review of 10000 alarms they would need to check just some 10 of them. And while there may be small amounts of "uncatchable" events that will be classified as false alarms by an operator but the fatigue and annoyance are incomparable.
The number of CCTV cameras connected to security operator centers or monitoring stations is growing exponentially. One operator handles events from many sites, which means that he/she has to observe dozens of cameras. The number of alarms generated by cameras can reach up to several dozen a day. Lack of proper filtering and classification of these events results in situations when either the human side becomes incapable of handling such a number of events and, consequently, the real alarms are lost in the clutter of other events. Or increases costs associated with the assignment of more and more employees required to handle this number of events. This is why Scylla's FAF comes to the rescue, which minimizes the number of false alarms to practically singles, allowing operators to focus on handling real alarms. And allows you to work with an ROI-effective solution that helps you to bring down costs of operations.
The whole point of the Scylla system is that it can be implemented in the most common video surveillance installations, with standard series of IPCs and NVRs. This means no special investment on the customer's part, and the mobility of the solution means that modules can be moved from channel to channel as required. In addition, using the Scylla solution means that the client's CCTV installations do not age out as quickly, so these funds can be invested in enhancing the security and safety of the facility by using other AI modules from Scylla.
By utilizing computer vision and artificial intelligence, Scylla False Alarm Filtering is designed to help security units by supporting their daily operations, augmenting their capabilities, and eliminating the overwhelming amount of false positives that inevitably cause noise fatigue, additional expenses, and time loss. For example, a security monitoring company can filter for false alarms, thereby reducing operator fatigue, and labor costs and increasing focus on investigating real issues. Another example: a factory has a problem with moths triggering security cameras that protect against human entry into dangerous areas. With false alarm filtering, these false alarms are eliminated, increasing operator focus on real security issues and perimeter intrusions.
An alert containing all the crucial information is compiled and delivered to end-users responsible for security. There are several customizable alerting pathways: Scylla dashboard, Scylla mobile alerting application, access point relay boards, and VMS alerting API.
Conventional motion detection-based surveillance cameras are inefficient for the most part as they trigger alerts each time anything moves on the video frame, be it foliage under the wind, shadow change, or an animal. Scylla AI video analytics understand what caused the motion and alerts are triggered only if the selected object is detected within the area of interest. Thus, it can operate in an outdoor setting.
False Alarm Filtering can snooze alerts by choosing the specified time period or just setting up a custom time range. False Alarm Filtering allows settings based on the site’s activity cycle.
Scylla has a powerful and advanced dashboard so you can easily follow all the locations and cameras and have total situational awareness. The Alarm Management dashboard allows you to see all alarms from all integrated devices with detailed information. You can see the snapshot of the detection occurrence and the recording of the alarm.
Scylla provides useful reporting to allow an analysis of alerts and to correct camera placement or security attention given to each area. Users can check and filter by date range the following statistics of each camera: Total alarms; Rejected alarms; Approved alarms; Historical timeline within the specified time range.
False Alarm Filtering detects humans and vehicles.
Yes, Scylla's FAF can work with any IP camera.
The event-based camera connection for integrations such as FTP, SMTP, and VMS is done through Scylla Cloud Dashboard, where you can find all connection requirements for needed integration types.
All alarms from all integrated devices with detailed information are available for monitoring from Scylla Cloud Dashboard.
By default, it is up to 100 users, but this limit can be increased upon request.