New possibilities of fire detection

Abstract

In this paper I explore new machine learning-based methods for detecting fires and technical rescues using visual information. My goal was to use artificial intelligence methods to extract information from street camera images that only humans could do before. This information was then used for fire protection purposes, i.e. automatic detection of damage. In my research, I used images downloaded from the internet to train a model to recognise fires and traffic accidents. I implemented the model in Python using the Pytorch framework. The model I created has become robust enough to automatically detect events with high confidence that require firefighter intervention.