Available online at www.ijournalse.org Emerging Science Journal (ISSN: 2610-9182) Vol. 8, No. 2, April, 2024 Page | 592 Comparison of Activation Functions in Convolutional Neural Network for Poisson Noisy Image Classification Khang Wen Goh 1 , Sugiyarto Surono 2* , M. Y. Firza Afiatin 2 , K. Robiatul Mahmudah 2 , Nursyiva Irsalinda 2 , Mesith Chaimanee 3 , Choo Wou Onn 1 1 Faculty of Data Science and Information Technology, INTI International University, Nilai, Malaysia. 2 Mathematics Study Program, Ahmad Dahlan University, Yogyakarta, Indonesia. 3 Faculty of Engineering and Technology, Shinawatra University, Pathum Thani 12160, Thailand. Abstract Deep learning, specifically the Convolutional Neural Network (CNN), has been a significant technology tool for image processing and human health. CNNs, which mimic the working principles of the human brain, can learn robust representations of images. However, CNNs are susceptible to noise interference, which can impact classification performance. Choosing the right activation function can improve CNNs performance and accuracy. This research aims to test the accuracy of CNN with ResNet50, VGG16, and GoogleNet architectures combined with several activation functions such as ReLU, Leaky ReLU, Sigmoid, and Tanh in the classification of images that experience Poisson noise. Poisson noise is applied to each test data to evaluate CNN accuracy. The data used in this study consists of three scenarios of different numbers of classes, namely 3 classes, 5 classes, and 10 classes. The results showed that combining ResNet50 with the ReLU activation function produced the best performance in class recognition in each scenario of the number of classes experiencing Poisson noise interference. The model achieved 97% accuracy for 3-class data, 95% for 5-class data, and 90% for 10-class data. These results show that using ResNet50 with the ReLU activation function can provide excellent resistance to Poisson noise in image processing. It was found that as the number of classes increases, the accuracy of image recognition tends to decrease. This shows that the more complex the image classification task is with a larger number of classes, the more difficult it is for CNNs to distinguish between different classes. Keywords: Activation Function; Classification; Convolutional Neural Network; Poisson Noise. Article History: Received: 23 November 2023 Revised: 26 February 2024 Accepted: 07 March 2024 Published: 01 April 2024 1- Introduction Deep Learning (DL) is one of the widely used technological tools because to its ability to represent images [1]. DL algorithms were designed with the aim of mimicking the function of the human brain, which has many hidden layers [2]. Deep learning (DL) algorithms have demonstrated notable success across diverse fields [3]. In the context of image classification, DL methods are often the first choice [4]. Another one of the most popular DL methods used for many image processing and human health applications is the convolutional neural network [5]. Convolutional neural network (CNN) is a machine learning algorithm that has shown to make good results for image classifications [6]. In recent years, CNN has attracted much attention due to its feature extraction capabilities and application to image classification [7]. CNN is an excellent classification method and is used widely today. One of the reasons is that CNNs provide high accuracy [8]. * CONTACT: sugiyarto@math.uad.ac.id DOI: http://dx.doi.org/10.28991/ESJ-2024-08-02-014 © 2024 by the authors. Licensee ESJ, Italy. This is an open access article under the terms and conditions of the Creative Commons Attribution (CC-BY) license (https://creativecommons.org/licenses/by/4.0/).