Comparison of Activation Functions in Convolution Neural Network

Abstract

The Convolution Neural Network (CNN) is a network with an input that is solely images. It is very useful as a powerful instrument for object recognition. This paper presents a part of a research in an area of the object recognition with a CNN for recognition of forest fires. The paper presents the different activation functions used in the CNN and the aim of the paper is a comparison between all of them. There are limitations in this field of research and in this paper information on them is provided.