AI-Driven fish disease diagnosis: A hybrid fusionNet19v4 approach

The cultivation of fish for human consumption in regulated settings is known as aquaculture. This rapidly growing sector within animal food production significantly contributes to the overall increase in fish output. However, diseases in fish pose major challenges, impacting fish health, production levels, and the sustainability of aquaculture practices. Factors such as stress, overcrowding, poor water quality, and insufficient quarantine measures for new or diseased fish contribute to various health issues. Therefore, early diagnosis and prevention of fish infections are essential to support the sustainability and well-being of fish farming. Addressing these issues through affordable, efficient tools for detecting fish diseases can enhance aquaculture outcomes. Numerous computer vision (CV) and machine learning (ML) techniques have been developed, though none have yet proven fully effective in diagnosing fish diseases. The proposed work introduces a machine learning (ML) and deep learning (DL) method to classify and diagnose fish diseases, including red spot and white spot conditions, ultimately leading to improved fish health management in aquaculture. This study uses a dataset of 1,364 images depicting red spots, white spots, and healthy fish. An ensemble learning strategy has been adopted to increase detection accuracy, with performance measured across two ensemble approaches, nine widely used classification algorithms, and performance evaluation matrices. The FusionNet19v4 ensemble model achieves a maximum accurateness of 96.64%, beating the pre-trained ResNet-50 model, which has an accuracy of 95.28%.