IMAGE-BASED CLASSIFICATION OF SIX PATHOGENIC BACTERIAL SPECIES USING EFFICIENTNET

M. Amirul Ghiffari El Ghiffari; Febri Dolis  Herdiani; Dea Aisyah  Rusmawati; Ishak Ariawan

doi:10.51876/simtek.v11i2.1839

Authors

M. Amirul Ghiffari El Ghiffari Industrial Technology, Politeknik Bhakti Asih Purwakarta, West Java, Indonesia
Febri Dolis Herdiani Information System, Institut Teknologi Al-Muhajirin, West Java, Indonesia
Dea Aisyah Rusmawati Department of Food Technology, University of Sultan Ageng Tirtayasa, Banten, Indonesia
Ishak Ariawan Marine Information System, Universitas Pendidikan Indonesia, West Java, Indonesia

DOI:

https://doi.org/10.51876/simtek.v11i2.1839

Keywords:

deep learning, EfficientNet-Lite0, image classification, pathogenic bacteria

Abstract

Image-based classification of pathogenic bacteria is a deep learning approach that can support automated bacterial identification. This study aimed to evaluate the performance of EfficientNet-Lite0 in classifying images of six pathogenic bacterial species, namely Bacillus cereus, Clostridium perfringens, Escherichia coli, Listeria monocytogenes, Salmonella typhimurium, and Staphylococcus aureus. The image data underwent preprocessing and augmentation during the training stage, followed by model evaluation using accuracy, precision, recall, and F1-score. The evaluation results showed that EfficientNet-Lite0 achieved 100% accuracy, precision, recall, and F1-score, with no misclassified images in the validation data. The model consisted of 3,378,694 parameters, with a computational complexity of 397.45 MFLOPs and a latency of 4.31 ms per image. These results indicate that EfficientNet-Lite0 achieved high classification performance for the six bacterial species under the dataset and evaluation conditions used in this study.