An Automated Fish Species Classification System Using Improved Alexnet Model

Due to the extremely dark nature of the sea's inner water and the fish's quick movement, classifying fish species from images obtained from the ocean presents significant challenges. This article describes an automated approach for identifying and classifying fish species using the deep learning method. It benefits marine scientists in various ways, most notably by allowing for the accurate monitoring of fish reproduction, development, and marine changes. AlexNet, a popular deep convolutional neural network model, is employed in this proposed study to classify fish species. This research modifies the traditional alexnet design to improve the accuracy of fish classification. In this proposed AlexNet architecture, five convolutional layers are used for an efficient texture and color feature extraction process. In addition, three fully connected layers are used for feature selection and classification. Finally, the classification efficiency of the proposed fish species classification system has been proven by comparative analysis with the most popular deep learning models (Alexnet, GoogleNet and VGGNet). The overall performance of the proposed deep learning model is 94%, its sensitivity is 95%, and its specificity is 95.5%, respectively.