Transit Terminals
Use views with a stable background where a plume is distinct from normal steam or glare.
Smoke in a terminal, concourse, or interior passage can be hard to notice across many views. As a secondary visual-detection layer, StreamVLM™ can route a visible plume to an operator for a rapid check; it does not replace a certified fire-alarm system.
The Visible Condition
The condition in scope is visible smoke, haze, or a plume in a selected interior camera view. It is a visible scene detail, not a diagnosis, a conclusion about intent, or an automatic decision.
In StreamVLM™, an authorized operator writes a prompt such as “Alert when smoke is visible in the terminal concourse.” The model evaluates selected live frames against that prompt and sends a match, camera, location, and time into the IREX event workflow for human review.
StreamVLM™ runs on a GPU node in the customer instance. Each prompt has its own confidence threshold, alert cooldown, event type, and camera selection.
Explore StreamVLM Prompt Detectors →Where It Helps
Use views with a stable background where a plume is distinct from normal steam or glare.
Have the control room visually verify the frame, adjacent cameras, and live scene promptly.
Hand a verified concern to the terminal’s established emergency-response process.
From Prompt to Action
Write a concise prompt for visible smoke, haze, or a plume in a selected interior camera view, then select the cameras and times where the condition warrants attention.
Open the event with its camera, location, time, and available video context. Confirm what is visible before treating the match as actionable.
The authorized team applies the local procedure. StreamVLM supplies a review cue; it does not decide the response.
Limits to Test
No. It is a secondary camera-based detection layer, not a replacement for certified fire-alarm or life-safety systems. Where smoke is visible to the camera, it may give an operator an earlier visual cue before smoke reaches a point detector or heat rises enough to activate a heat detector. The operator still visually verifies the alert and follows the facility’s established emergency procedure.
No. It reports a match to the visible condition in the prompt. An operator reviews the frame and local context before deciding what it means.
Start with representative cameras, define what makes a useful alert, then adjust the camera selection, confidence threshold, and alert cooldown during a camera-specific pilot.
IREX does not publish a universal accuracy figure for this prompt. Image quality, camera angle, lighting, occlusion, and the local definition of a useful alert all affect results.
Start with one concourse view where normal steam, haze, and lighting conditions are known to the operations team.