Fish Classification with Machine Learning: Enhancing Accuracy and Efficiency

Abstract

Fishery sector is regarded as a sunrise sector and is expected to play a significant role in the near future. This paper emphasizes the need for a comprehensive aquaculture supervisory system that prioritizes fish identification and classification in order to effectively monitor and control a variety of aquaculture operations. In order to automatically identify and classify fish species from a variety of data sources and do away with manual identification, the suggested approach integrates Artificial Intelligence andMachine Learning technologies, including CNN, MobileNetV2, ResNet152, and YOLOv8 models. Fish image analysis using deep learning and other cutting-edge methods improves accuracy by identifying complex patterns. This approach is meant to be adaptive and flexible; it can be changed in response to new facts and circumstances. A fruitful implementation would boost fish identification and classification for effective management, as well as aquaculture's sustainability, profitability, and efficiency. With a minimum loss function of 0.02 and an accuracy of 94.8%, YOLOv8 stood out for its exceptional performance, demonstrating its potential for high-accuracy testing and training in the context of fish identification and classification techniques. However, to fully realize the benefits of AI and ML in aquaculture, issues like the scarcity of high-quality training data and the requirement for specialized knowledge in these fields must be resolved.