The Challenge of Aqua Creatures Specie Classification in the Aquarium Innovation Theme

Abstract

The aqua creatures are lively, diverse, and abundant; nowadays, their survival depends on the proactive endeavor of human society on the environment cares; recognizing them conveniently and effectively is the first step to care. Recently, the fast-growing applications of the computer vision techniques along with the Internet-of-Things attract both the researchers’ and the practitioners’ attention. Such applications give the alternative to the traditional approaches to observe the moving objects more efficiently with higher precision through image capturing. In the common aquarium themes, the compartment may contain the same aqua-creature specie or non-mutual offensive species. The objective of the mono-specie scenario identification is to tell the difference between the compartments’ species, while the other scenario can identify the specie of the individual aqua-creature within the multi-specie compartment. This paper is the few studies that aim to facilitate the aquarium operations, especially in the animal state observation. It discusses the technical challenges in dealing with the aqua-creatures images collected from the aquarium scene. For this purpose, the paper presents two comprehensive aqua-creature identification approaches were applied the neural
networks for different operational scenarios. The contribution of this paper is to explore the potential in caregiving operations and the aquatic education based on the computer vision techniques of species identification. Further derived applications, such as illness detection and adult-creature counting, can be widely applied in the real aquaculture farm.