Jurnal Elektronika dan Telekomunikasi (JET), Vol. 23, No. 1, August 2023, pp. 47-54 Accredited by KEMDIKBUDRISTEK, Decree No: 158/E/KPT/2021 doi: 10.55981/jet.533 Bacterial Classification Using Deep Structured Convolutional Neural Network for Low Resource Data M Faizal Amri a,* , Asri Rizki Yuliani b , Dwi Esti Kusumandari a , Artha Ivonita Simbolon a , M. Ilham Rizqyawan a , Ulfah Nadiya a a Research Center for Smart Mechatronic, National Research and Innovation Agency Jl. Sangkuriang, Dago Bandung, Indonesia b Research Center for Artificial Intelligence and Cyber Security, National Research and Innovation Agency Jl. Sangkuriang, Dago Bandung, Indonesia Abstract Bacterial identification is an essential task in medical disciplines and food hygiene. The characteristics of bacteria can be examined under a microscope using culture techniques. However, traditional clinical laboratory culture methods require considerable work, primarily physical and manual effort. An automated process using deep learning technology has been widely used for increasing accuracy and decreasing working costs. In this paper, our research evaluates different types of existing deep CNN models for bacterial contamination classification when low-resource data are used. They are baseline CNN, GCNN, ResNet, and VGGNet. The performance of CNN models was also compared with the traditional machine learning method, including SIFT+SVM. The performance of the DIBaS dataset and our own collected dataset have been evaluated. The results show that VGGNet achieves the highest accuracy. In addition, data augmentation was performed to inflate the dataset. After fitting the model with augmented data, the results show that the accuracy increases significantly. This improvement is consistent in all models and both datasets. Keywords: Bacterial classification, Deep learning, Convolutional neural network (CNN), E-coli. I. INTRODUCTION Bacteria are living microorganisms that are not visible to the naked eye but can be observed under a microscope. While some bacteria benefit our ecosystem, they can also become a carrier of diseases to humans, including infectious ones. It is, therefore, essential to provide accurate identification of living bacteria for a wide range of applications, including clinical diagnosis [1], [2], food production [3], and water quality assessment [4], [5]. The characteristics of bacteria can be examined under a microscope using culture techniques. However, traditional clinical laboratory culture methods require considerable work, primarily physical and manual effort. The recognition and counting of bacteria colonies are mainly performed by the expert. Moreover, some bacteria species belong to the same morphology, making identifying them more difficult. To cope with these problems, an automated identification and classification process can be developed. Machine learning (ML) assisted image processing has been comprehensively used to reduce the workload and improve performance accuracy. Specifically, deep learning (DL) as a subset of ML has achieved incredible success in a number of biomedical applications [6], [7]. Deep learning approaches are able to learn features from large datasets to reach human-level performance. Recently, many studies have been developing an automated system using DL that can assist biologists and related researchers in recognizing microscopic images in large-scale applications, not only bacteria identification but also other living microorganisms. In [8], they provided a systematic review of various ML methods applied for image recognition of four types of microorganisms such as bacteria, algae [9], protozoa [10], and fungi [11], [12]. Other than that, [13] provided a comprehensive survey of digital image processing methods for microorganism counting. The study concluded that there is an urgent need for the automation of bacteria colony detection with high accuracy and acceleration. Most studies usually work with an adequate number of images. However, collecting bacterial data is labor-intensive and error-prone. To provide a highly accurate system, this paper presents a deep structure convolutional neural network (CNN) for bacteria classification when low resource data have occurred. The utilization of CNN was performed because it has become state-of-the-art for many object detection and classification tasks. The first step in performing our methods was to design a standard CNN architecture with three convolutional layers as a baseline. In the next step, the baseline results were compared with gated CNN and two popular pre-trained models based on * Corresponding Author. Email: mfai001@brin.go.id Received: January 19, 2023 ; Revised: March 31, 2023 Accepted: Apr 26, 2023 ; Published: August 31, 2023 Open access under CC-BY-NC-SA © 2023 BRIN 25-32