Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Jurnal Teknologi Informasi, Komputer, dan Aplikasinya (JTIKA )
Maftuh Ahnan Al-Kautsar
Department of Informatics Engineering, Faculty of Engineering, University of Mataram
Computer Science & IT Decision Sciences, Operations Research & Management Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
KLASIFIKASI UNTUK MENDETEKSI PENYAKIT IKAN MENGGUNAKAN MODEL VGG-16 DAN MOBILENETV2 Maftuh Ahnan Al-Kautsar; Fitri Bimantoro; I Gede Pasek Suta Wijaya
JTIKA (Jurnal Teknik Informatika, Komputer dan Aplikasinya) Vol 8 No 2 (2026): September 2026
Publisher : Program Studi Teknik Informatika, Fakultas Teknik, Universitas Mataram

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29303/jtika.v8i2.661

Abstract

Fish disease is a major challenge in the aquaculture sector, causing global economic losses of up to US$6 billion annually. Conventional disease identification relies on visual observation and laboratory testing, which is time-consuming, costly, and requires specialized expertise. This study investigates freshwater fish disease classification using two Convolutional Neural Network (CNN) architectures based on transfer learning, namely VGG-16 and MobileNetV2, combined with sequential hyperparameter tuning. The dataset consists of 2,800 images across eight classes, comprising 2,000 training images and 800 images used for validation and final evaluation. Sequential tuning was performed across five hyperparameters: optimizer, learning rate, dropout, batch size, and number of epochs. The best configuration of VGG-16 (Adam optimizer, learning rate 1e-4, dropout 0.2, batch size 16, and 50 epochs) achieved an accuracy of 99.62%, while MobileNetV2 (Adam, learning rate 1e-3, dropout 0.2, batch size 16, and 50 epochs) achieved an accuracy of 99.37% on the evaluation data. VGG-16 produced a slightly higher accuracy, with a difference of 0.25 percentage points, whereas MobileNetV2 had substantially fewer base-model parameters. These results indicate that both architectures achieved high classification performance under the experimental setting used in this study, while their differences in model complexity may be considered when selecting a model for further deployment. However, the use of the same data for validation and final evaluation represents a limitation that may result in an optimistic estimation of model performance
Co-Authors Fitri Bimantoro I Gede Pasek Suta Wijaya