Artificial Neural Network Techniques for Differentiating Poisonous and Edible Fish Species

Abstract

Oman is known for its diverse marine life, including many species of fish that can pose a risk to human health if eaten. In this study, we aim to develop an effective system that classifies between edible fish and poisonous fish based on the combination of CNN-MLP networks to identify poisonous fish. A hybrid architecture that incorporates convolutional neural networks (CNNs) will be utilized for feature extraction and multilayer perceptrons (MLPs) for additional descriptive features. The model will be trained with a combination of CNN-MLP architecture using backpropagation to improve prediction accuracy. Our model achieved a 92.4% accuracy rate. The results of this study will contribute to the understanding of toxic fishes in Oman and will provide a valuable tool to mitigate risks associated with marine life in the region and enhance safety.