IREX Launches StreamVLM for Custom Public Safety Video Analytics

IREX, an ethical AI company based in California, has launched a groundbreaking new tool called StreamVLM™. This Vision-Language Model (VLM) detection engine allows public safety operators to create custom video analytics detectors using plain-language descriptions, eliminating the need for complex dataset collection and model training.

Traditionally, adding a new capability to a video analytics system required a lengthy and resource-intensive process: collecting a dataset, labeling it, training a model, validating it, and finally deploying it. This cycle could take months and significant budgets. With StreamVLM, the process is dramatically simplified.

Users can now describe what they want the system to detect using ordinary English. For example, they can type in a sentence like, "Detect a person lying on the ground," "Alert when graffiti appears on a wall," or "Identify flooding in the underpass." The platform then begins monitoring selected cameras for these conditions in real-time.

Serge Smirnoff, IREX's Head of PR, explained, "Public safety agencies have never been short on things they need to see. They have been short on time and money to build a model for each one. StreamVLM changes who gets to decide what a camera network watches for. It is no longer a data science project; it is a sentence typed by the person who actually knows the neighborhood, the station, or the campus."

A StreamVLM detector is a named set of prompts with its own settings, applied to selected camera channels. Each prompt is independent, with its own confidence threshold, alert cooldown, and event type. Selected frames from live camera feeds are evaluated continuously against the prompts by a VLM that understands images and language together. Matches generate real-time alerts within seconds.

A single camera channel can support multiple prompt-defined detectors simultaneously. For instance, a station camera can watch for a person on the tracks, platform overcrowding, an unattended bag, smoke, fresh graffiti, and flooding at the platform level. Detectors can be added or adjusted at any time without taking the system offline.

StreamVLM is part of IREX's broader suite of specialized analytics modules, which include facial recognition, vehicle and traffic analysis, weapon detection, perimeter security, rail and transit monitoring, crowd management, fire detection, and camera integrity checks. However, it extends these capabilities into areas not covered by fixed module catalogs, such as infrastructure damage, illegal dumping, snow and ice hazards, unattended objects, unauthorized vehicles, worksite safety violations, non-standard signage and vehicle markings, and conditions particular to a single city.

According to Smirnoff, "Every prompt is logged, attributed, and reviewable, ensuring the same level of oversight as a custom-trained model. The same mechanisms that govern the rest of the platform are in place to ensure transparency and accountability."

The release of StreamVLM marks a significant advancement in the field of public safety technology, offering a more efficient and flexible solution for monitoring video feeds. With its ability to quickly and easily create custom detectors, StreamVLM is poised to revolutionize the way public safety agencies operate.