Neural Networks Fire Detection in a Public Safety Fog System

This paper presents a fire detection method based on neural networks used for a Flying Ad-Hoc Network (FANET) that is part of a Public Safety Fog System Architecture. The complete architecture of this solution and the interconnection between all of the components is depicted in this work. A five hundred pictures database was build and TensorFlow from Google was used to model and train the neural network. The performance evaluation of the implemented algorithm is provided and discussed. Based on previous results, the fire detection mechanism is the last required validation step towards a real prototype implementation of the proposed system model.