Fire Detection In Development Expected by end of Q4
What it does:
Looks for the horizontal shape of a fire or smoke in video feeds.
Allows you to create alerts, filter event feeds and 24/7 views based on the presence of a fire or smoke.
Training Data Limitations
Candles, fireplaces, or bonfires are fires and will be detected as such. It is not trained to and cannot tell the difference between a managed and an out-of-control fire. It is recommended that you use exclusion zone(s), if candles, fireplaces, or other intentional fires are in view.
Consensus Based Architecture Expected in Q4
The fire detection model uses our consensus machine learning architecture which means that you will receive a report not on one computer vision model but 2 or 3 (in the case of car fires). Each analytic will report its own estimate on the likelihood of the event being a fire.
These models were trained on horizontal fires. It is not intended to recognize fire from a top-down point of view. It is trained on indoor and outdoor fires, including car fires.
Enabling Fire Detection requires its own EULA that limits our liability. Fire detection is obviously not a replacement for a fire detector or fire suppression system and will not meet the regulatory requirements for these devices.
This model was made by Survail and will continue to learn.