IREX Launches StreamVLM™: A Game-Changer in Public Safety Analytics
In an inspiring move towards enhancing public safety, IREX, a company devoted to ethical AI practices, has recently introduced its innovative
StreamVLM™, a cutting-edge Vision-Language Model (VLM) detection engine. This groundbreaking technology allows public safety operators to convert plain language descriptions into efficient video analytics detectors, representing a first of its kind in the industry.
Revolutionizing Surveillance with Plain Language
Traditionally, adding new functionalities to video analytics systems involved a lengthy process that included commissioning new AI models, collecting datasets, and training the models, a venture that often stretched over several months and required significant financial investment. However, IREX’s StreamVLM™ dramatically streamlines this process. By simply describing the desired detection in natural English, such as "identify flooding in the underpass" or "alert when graffiti appears on a wall," the platform can immediately begin monitoring the selected cameras for these conditions.
According to
Serge Smirnoff, IREX's Head of PR, public safety agencies often face an overwhelming amount of information and threats to monitor but lack the budget and time to deploy extensive model training for every single requirement. With StreamVLM™, the power shifts to the operators who are familiar with the specific environments they oversee, allowing them to dictate exactly what the camera networks should observe.
Technical Insight into StreamVLM™
The StreamVLM™ operates across pre-existing camera networks, whether from a single site to nationwide deployments. This capability enables public safety agencies to conduct intricate investigations using natural language and real-time threat detection through specialized AI video analytics modules. The adaptability of the system is significant, as it allows event creation, editing, and activation, with every action being logged under the operator’s account for accountability and review.
Detectors are essentially sets of prompts that can be independently customized for multiple video channels. For instance, a single camera can monitor for various conditions, such as rail safety by detecting people on tracks, detecting unattended bags on platforms, or identifying overcrowding in real time. The flexibility to add or adjust detectors on the fly without system downtime is a game-changer in public safety surveillance.
Furthermore, StreamVLM™ works seamlessly alongside existing IREX analytics tools, which cover various parameters including faces, vehicles, and traffic patterns. However, it also fills the gaps left by conventional modules, extending its capabilities into previously unaddressed areas such as monitoring for infrastructure damage, illegal dumping, or hazardous conditions unique to specific locales.
Governance and Oversight
Despite the simplicity of language commands, each detector created through StreamVLM™ comes with the same rigorous oversight as traditional models. Every prompt is recorded, and all events are auditable. The system allows for role-based access, ensuring that investigations are meticulously tracked and managed under a unique Case ID.
This remarkable innovation not only changes the landscape of video analytics for public safety agencies, but also democratizes the process, giving control back to those who best understand their communities’ needs. By empowering operators to define observance criteria without reverting to complex data science, StreamVLM™ signifies a monumental shift towards smarter and more responsive public safety strategies.
Conclusion
With the introduction of StreamVLM™, IREX is poised to revolutionize how public safety organizations utilize video surveillance. By fostering a user-driven analytics environment, the company remains at the forefront of ethical AI technology that prioritizes safety, adaptability, and community understanding.
Watch the demonstration here and see how IREX is setting new standards in the public safety landscape.