Future Trends and Short-Review on Fish Species Classification Models Based on Deep Learning Approaches

An automatic fish species classification system is highly essential to track, classify and detect marine species and fishes in complex underwater backgrounds without any manual help. Existing computer vision-based approaches didn't offer effective performance rates underwater because the textural features and shape of fish species are not apparent. The data-driven classification approach needsa superior amount of labeled data;otherwise, they leadtooverfitting at the time of data training, and also, the unseen test data are not utilized for training.Ecologists, researchers, taxonomists, and geneticists from different biological fields wished to accept fish as a significant element in their research. Theywere discouraged from finding ichthyology,a highly complex task. In fish-based studies, failing to recognize fishes as distinct biological units can lead to thewrong diagnosis. Hence, this survey paper tries to discuss and clarify the recent research on species identification with their algorithmic categorization. This review explores the performance measures compared, the number of fish families/species recognized, datasets used, and tools utilized for implementation. Further, future research directions and compensation for current research gaps are discussed. This review of state-of-the-art fish identification tools shows their potential for providing the right solution in real-life situations.