Manuscript received March 28, 2025; revised April 29, 2025; accepted June 30, 2025; published January 16, 2026.

Abstract—The assessment of fish freshness is crucial for ensuring food safety and quality within the seafood industry. Traditional methods of freshness evaluation rely on sensory and instrumental assessments, which can be subjective and require expertise. This study explores the application of machine learning techniques to classify fish freshness based on eye images. A total of 880 images were collected from two fish species—milkfish (Chanos Chanos) and tilapia (Oreochromis Niloticus)—with distinct eye characteristics, spanning four freshness categories: excellent, good, average, and not fit for consumption. Features extracted from the eye regions, including RGB, CIE Lab*, and GLCM descriptors, were used to train three classification models: Naïve Bayes (NB), Support Vector Machine (SVM), and k-Nearest Neighbors (KNN). Among the models, KNN achieved the highest accuracy of 77%. The study demonstrates the potential of automated, non-destructive, and objective machine learning-based approaches for evaluating fish freshness, contributing to improved quality control in the seafood industry.

Keywords—fish freshness, image processing, k-Nearest Neighbors, machine learning, Naïve Bayes (NB), seafood quality, Support Vector Machine (SVM)

Cite: John P. Q. Tomas, Maria R. J. C. Caranay, Eonn G. M. G. Domingo, Mia B. Enciso, and Bonifacio T. Doma, "Application of Machine Learning to Determine Fish Freshness Based on Eye Images," Journal of Image and Graphics, Vol. 14, No. 1, pp. 24-37, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.