Skip to content

Here we present the trained model and the algorithm that let us win the ICIAP2023 ONFIRE contest

License

Notifications You must be signed in to change notification settings

AI-FIREBUSTERS/ICIAP_ONFIRE

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 

Repository files navigation

ICIAP ONFIRE COMPETITION

The Problem

Real-time fire detection from image sequences is a highly requested feature in real video surveillance applications, as it can prevent environmental disasters and ensure continuous monitoring of the urban environment and the protection of the forest heritage. In this scenario, there is a great interest to cover the territory with "intelligent cameras" with on board video-analytics algorithms able to detect fires (flames and/or smoke) in real time, generating immediate notifications for the alarm centers of the competent authorities; these cameras, typically installed in isolated environments, must be self-sufficient from a computational point of view, at most accompanied by a small embedded system that processes the sequence of images by applying the fire detection algorithm. Therefore, very effective methods that require a huge amount of processing resources are useless in this application; it is necessary to find a good trade-off between fire detection accuracy, notification promptness and processing resources.

The Contest

The ONFIRE 2023 contest is an international competition among methods, executable on board of smart cameras or embedded systems, for real-time fire detection from videos acquired by fixed CCTV cameras. To this aim, the performance of the competing methods will be evaluated in terms of fire detection capabilities and processing resources. As for the former, we consider both the detection errors and the notification speed (i.e., the delay between the manually labelled fire start, either its ignition or appearance on scene, and the fire notification). Regarding the latter, the processing frame rate and the memory usage are taken into account. In this way, we evaluate not only the ability to detect fires and avoid false alarms of the proposed approaches, but also their promptness in notification and the computational resources needed for real-time processing. You can find more info on OnFire Contest 2023

About

Here we present the trained model and the algorithm that let us win the ICIAP2023 ONFIRE contest

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published