2 min read
Introduction
Specialized Object Detection (S.O.) is not just a new name for weapon detection – it is a comprehensive algorithm designed to enhance Vaidio’s accuracy with complex objects. It identifies small, complex objects of interest * when they are associated with another object or person * and is particularly useful when high accuracy is crucial.
Core Functionality: Object Association
At the heart of S.O. is the concept of object association. For example, rather than just identifying a gun on a table, S.O. is designed to alert when a gun is held by a person. This association is central to the operation of S.O., enabling it to avoid false positives (like alerting for a gun on a table when that isn't deemed a threat) and to deliver higher accuracy on significant events (like a person holding a weapon).
Integration with Object Detection
S.O. does not work in isolation; it operates in conjunction with Vaidio’s general object detection model. Every Vaidio engine, including S.O., requires the foundational layer of object detection. While object detection is a part of the package when licensing S.O., it operates as a separate layer, providing the necessary groundwork for the more specialized analytics of S.O.
Technical Insights: Complex Object Processing
The S.O. engine employs a unique process. Initially, it uses general object detection to identify a primary object (e.g., a person) in the video frame. If this primary object is detected, the system then applies S.O.’s unique reprocessing to identify the secondary, more complex object of interest (e.g., a weapon or a cell phone). This approach allows for enhanced accuracy and reduction in false positives, even relative to competitive point products.
Licensing
For the Vaidio 7.1 release, S.O. will be in beta in order to gain real-world feedback and to tune the model. This initial release will enable weapon detection only. (We intend to add additional complex and associative objects such as cellphones, tools, handheld radios and etc. in future releases.) Like every Vaidio analytic, S.O. will come bundled with object detection and with forensic video search. And, like every Vaidio analytic, Specialized Object Detection is licensed separately and per camera.
Application
S.O. is intended for situations where high accuracy is paramount, especially when objects are small and/or can be easily confused with other items under certain conditions. For example, a cell phone or a tool might be mistaken for a handgun under certain angles and lighting conditions. In these scenarios, S.O. is beneficial because it adds an additional layer of verification to the detection process and distinct advantages in terms of accuracy.
The Balance of Cost and Accuracy
Integrating S.O. does come with additional costs, as it essentially doubles the compute resource requirements compared to normal object detection. This makes it essential for integrators to weigh the cost against the need for increased accuracy. In general, S.O. is recommended when the object of interest is easily confused with another item, or when pinpoint accuracy is absolutely critical.
Conclusion
Specialized Object Detection (S.O.) represents a significant advancement in Vaidio video analytics, offering a sophisticated solution for scenarios demanding high accuracy. By focusing on the association between objects, S.O. provides a more targeted and effective detection process, making it a valuable tool for security and surveillance applications. Learn more by downloading the Specialized Object Detection datasheet below.
