Дякуємо, ми отримали ваше повідомлення і звʼяжемось в найближчий час! :)
EN
EN
UA
EN
UA
How It Works:
● Scylla is optimized to handle multiple video streams on a single GPU, enabling real-time event tracking.
● Upon recognizing an anomalous event based on a series of frames provided to the model, Scylla sends alerts to all designated endpoints.
● The module supports real-time multithreaded processing as well as offline video analysis.
● Models are trained on large datasets of both anomalous and typical video recordings. This enables Scylla to operate effectively across diverse environments and scenarios, providing immediate alerts upon anomaly detection.
● The system continuously self-learns, allowing customization for your specific use case if the current dataset does not fully capture the unique characteristics of your environment.
What Does Scylla Anomaly Detection and Behavior Recognition Identify?
The system is trained to detect and identify a wide range of suspicious behaviors:
Vandalism detection
Aggressive behavior detection
Slip and fall detection
AI-Based Technologies for Public Safety and Risk Reduction
Robbery Mask Detection
In situations where individuals attempt to conceal their identity during a robbery using masks or balaclavas, traditional surveillance methods may fail to provide timely and accurate threat assessment. The automatic detection technology utilizes advanced algorithms to analyze video streams in real time, identifying the presence of robbery masks or balaclavas.
Foreign Object Debris
Foreign object debris at airports can cause significant damage, costing airlines and airports millions of dollars annually. Such debris can be found near terminal gates, cargo aprons, runways, and other airport locations. It is estimated that FOD-related damages cost the aerospace industry around $4 billion per year, not to mention potential aircraft damage and injuries.
Scylla’s AI-powered object detection can effectively analyze video from high-quality cameras, detect various unattended objects at a distance, and send alerts about possible foreign object debris to all designated endpoints.
Detection of Abandoned Objects
Detection of abandoned objects (AOD) is a critical issue as transportation hubs, critical infrastructure facilities, and law enforcement agencies aim to utilize their video surveillance camera networks to identify unattended bags in public spaces both in real time and for forensic analysis. Abandoned items pose a potential threat to public safety and require immediate attention from police and security personnel.
Scylla’s proprietary object detection algorithms can identify abandoned objects and notify security staff as soon as they are detected. This enables timely action and saves valuable response time.
Trash Detection
Trash in public places, on roads, and at transportation hubs not only poses environmental hazards but can also lead to property damage.
Most of these items are difficult to identify using conventional artificial intelligence. However, Scylla employs robust object detection algorithms that enable it to detect small unattended objects at a distance, even against moving backgrounds.
What distinguishes the Scylla Object Detection and Tracking system from other solutions?
● AI methodologies operate autonomously 24/7 and continuously improve themselves.
● Easily integrates with most cameras and video surveillance systems.
● Works effectively with cameras featuring moving backgrounds, such as drones and body-worn cameras.
● Utilizes a built-in zoom and tracking algorithm that enables detection of distant objects.
● Requires 7 times less hardware resources compared to similar solutions on the market.
● Can be deployed both locally and in the cloud.
● The object tracking module is based on a proprietary algorithm that is lightweight, accurate, and versatile.
● The AI model is custom-designed for re-identification of specific object types with unparalleled precision.
● The algorithm can be easily applied in centralized systems that analyze hundreds of video streams simultaneously.
Fill out the form to receive a personalized consultation for your IT infrastructure:
FAQ
Can the Scylla Object Detection system effectively operate outdoors?
Yes, the systems are trained on a wide variety of landscapes and backgrounds, under different lighting conditions and from various angles. Essentially, the solution is quite resilient to background variations — as long as the object of interest is visible, the system will reliably detect it.
What is the maximum detection distance?
The answer depends on several factors, primarily the camera’s characteristics, especially its resolution. Resolution initially plays a major role; however, Scylla’s built-in scaling and tracking algorithm allows the system to analyze the object at the camera’s original resolution. Unlike similar AI-based security solutions, Scylla is less dependent on the visual image quality, which typically degrades when processed through neural network platforms.
There are a number of factors related to the camera’s "image quality," such as stream bandwidth, encoding, etc. Visibility conditions—like lighting, object orientation (see question 6 regarding object angle), and pixel size of the object—also need to be considered. The latter depends linearly on the distance from the camera and can be used to estimate the maximum detection range.
For example, a reliable minimum object size for firearms detection is about ~15-17 pixels, which corresponds to a maximum distance of approximately 10-12 meters for most HD cameras.
What is the detection response time?
Detection typically occurs within the first 400 ms (in some cases, up to 2 seconds). When evaluating response time, keep in mind that most IP cameras used today introduce some sub-second video stream latency. Additionally, in cases where Scylla is deployed in the cloud, you should consider the time delay caused by the stream reaching the cloud and the response being delivered to the dashboard.
How does the Scylla Object Detection System assist law enforcement?
The Scylla Object Detection System is designed to support security units in their daily operations by enhancing their capabilities and eliminating potential shortcomings related to human error. Additionally, in the event of a potential threat, the alerts generated by Scylla are enriched with critical information that enables rapid and comprehensive on-site threat analysis and effective planning of countermeasures.
Can it detect a weapon if someone is carrying it in a bag?
The system is based on computer vision algorithms, and threat detection relies on visual content analysis. This means that to detect a weapon inside a bag, Scylla must be connected to X-ray or millimeter-wave scanners. When operating on surveillance cameras that work solely in the visual spectrum, the Scylla Object Detection System can only detect visible, unconcealed weapons.
Does the object need to be at a specific angle for the system to detect it?
No, the system is trained to recognize objects from all possible angles. However, in certain specific cases, the angle of the object may influence detection accuracy, as the features Scylla uses to classify an object can be more distinct from some angles than others. For example, if a gun or rifle is held at an angle to the camera, it may reveal more distinguishable features compared to when it is pointed directly at the camera.
What does the system do after detection?
An alert containing all relevant information is generated and sent to the end users responsible for security. There are several configurable notification channels, including the Scylla Dashboard, the Scylla mobile app, access point relay boards, and VMS alert APIs, among others.
Can Scylla Object Detection operate in the cloud?
Yes, all Scylla solutions can be deployed both in the cloud and on-premises. Moreover, Scylla's AI-powered software solutions are cloud-provider agnostic, as long as the cloud instance runs on Linux and is equipped with an Nvidia GPU.
Can the Object Detection System work with infrared/thermal cameras?
Yes, it can. The maximum detection range of the Scylla Object Detection Solution depends on the specifications of the camera, such as contrast ratio, pixel crosstalk, and other factors. However, in general, the solution complies with standard industry DRI requirements—meaning the identification threshold (the distance at which the class of an object can be determined) is approximately 20 pixels for small arms.
What are true and false alarms?
An alarm is classified as true when the AI’s prediction aligns with reality (i.e., the object of interest is correctly identified, or the intended action is accurately detected). A false alarm occurs when an alert is triggered incorrectly. Unfortunately, due to the probabilistic nature of AI, false alarms are largely unavoidable. However, thanks to the advanced artificial intelligence and machine learning powering the Scylla Object Detection System, it meets any industrial-grade standard. Moreover, Scylla AI’s video analytics modules are continuously improved through retraining on past errors, further reducing the number of false positives over time.
What are the recommended camera parameters?
Most questions regarding camera limitations and requirements can be answered with a simplified “empirical” rule: if a person can see and identify an object of interest, then the Scylla AI video analytics system will also be able to do so (and in some cases even surpass a person thanks to built-in scaling and re-checking algorithms). As for the minimum camera parameters, they depend on each specific use case and the object you are trying to detect. Of course, the camera must have a digital output or, at least, be connected to a video recorder that has one. The Scylla object detection system can accept virtually all types of streams, such as RTSP/RTMP, HTTP, etc. Typically, the minimum required resolution starts at HD (1280x720) and 5 FPS. The parameters that determine frame/image quality vary from camera to camera, but we recommend paying attention to characteristics such as bandwidth, encoding, and sharpness, and improving them if necessary.
Translated with DeepL.com (free version)
Does the Scylla Object Detection System comply with GDPR and CCPA requirements?
Absolutely. Scylla does not store any information unless explicitly requested by the user.
Can the solution operate in harsh conditions?
The Scylla Object Detection System is designed to function effectively in challenging environments where cameras with built-in algorithms may fail to perform properly. The AI engine compensates for limitations caused by difficult conditions, including poor lighting, slightly distorted frames, environmental factors, and weather effects.