Classification of Freshwater Broodstock Using Image-Based Machine Learning

The extinction of freshwater fish is mainly due to environmental destruction, which has led to an increase in fish breeding activities in the aquaculture sector. This study presents the implementation of a machine learning approach using the YOLOv5 model to facilitate the fish breeding process by classifying the sex of Mahseer (Kelah) and the gravid broodstock of Tilapia. Traditional methods of sex and broodstock identification are often time-consuming and prone to errors due to a lack of expertise and equipment. A collection of X-ray images of Tilapia fish was gathered and labeled to develop a dataset for training and evaluating the YOLOv5 model. Due to the high cost of obtaining original data for Mahseer, publicly available images from Google were used to build the dataset. The results demonstrated that the YOLOv5 model achieved a precision of 94% and 99.4% for classifying sex and broodstock, respectively, using a maximum dataset of 430 images and 40 images. This finding indicates that YOLOv5 can significantly improve the efficiency and accuracy of broodstock management in aquaculture, contributing to the growth of the aquaculture sector.