Fish image classification is seemingly simple yet a convoluted process. Moreover, the scientific research of population counts and geographical behaviour is substantial for progressing the current developments in this field. We've tried several approaches to find the optimum-performing approach using advanced computer vision and data mining techniques with limited research scope and difficulties. Its performance was compared to the state -of-art models like CNN, EfficientNet etc., to validate the credibility of the proposed model. Eventually, it was observed that the empirical approach using the ANN confirmed the DNN model to be the leading model with an accuracy of 100%.