Deep Learning for Fire and Smoke Detection in Outdoor Spaces

Abstract

AbstractRecent investigations in deep learning provide a new approach for fire and smoke detection in outdoor spaces. Fire and smoke detection is an additional function in smart urban surveillance, as well as, in wildlife monitoring using stationary cameras or UAV cameras. Smoke detection plays an important role in a fire alarm. The algorithms based on the traditional machine learning techniques show high values of errors that cause additional economic costs. Deep learning techniques solve this problem partly but raise new challenges. In this chapter, we analyze deep learning models applicable for this task and present a Weaved Recurrent Single Shot Detector (WRSSD) for early smoke detection with acceptable error ratios.KeywordsFire detectionSmoke detectionDeep learningOutdoor space