Citation

Shaik Mohamed, Muhammad Faris and Yunos, Muhammad Yazid and Ali, Aziah and Hashim, Noramiza (2026) Identification of Fish Species from Local Fish Market Images. JOIV : International Journal on Informatics Visualization, 10 (3). p. 1247. ISSN 2549-9610

Text 4206-15503-1-PB.pdf - Published Version Restricted to Repository staff only Download (4MB)

Abstract

In Malaysian local markets, identifying fish can be challenging, especially for consumers unfamiliar with various species. However, recent technological advancements in machine learning, specifically deep learning, have rapidly gained prominence across various industries, including aquaculture. When applied to fish identification, deep learning faces both challenges and exciting possibilities. This research focuses on developing a fish species identification system using images from local fish markets. The methodology involves several steps, including data preprocessing and augmentation, which enhance the dataset's diversity and improve model training. But several concerns need to be addressed. First, the lack of comprehensive local market datasets in Malaysia poses a challenge. Additionally, research on fish species identification within these markets remains limited. The complexity arises from identifying fish species under varying environmental conditions, different lighting, and scenarios involving multiple fish. In this study, the dataset comprises 4500 images representing 15 species. Several deep learning models, namely DenseNet121, VGG16, and MobileNetV2, are implemented to evaluate their effectiveness for fish species identification. The results indicate that DenseNet121 achieved the highest accuracy, with a remarkable test accuracy of 99.78% and a low-test loss of 0.01%. Data preprocessing steps, such as image augmentation and resizing, played a crucial role in improving model accuracy. The findings underscore the importance of expanding the dataset to include a broader range of fish species, which could further enhance model accuracy and generalization capabilities.

Item Type:	Article
Uncontrolled Keywords:	Deep learning
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	30 Jun 2026 08:46
Last Modified:	30 Jun 2026 08:46
URII:	http://shdl.mmu.edu.my/id/eprint/16164

Downloads

Edit (login required)