Analysis of Fish Behavior Using Deep Learning Techniques

Aquaculture plays a crucial role in the fishing industry, ensuring both food security and sustainable recovery of biological resources. It aims to provide affordable and high quality fish products, which contributes to maintaining ecosystem stability and economic development of Russia's fisheries sector. To solve this problem, specialized aquaculture complexes and production facilities are being created, the main purpose of which is to regenerate and support fish populations. Such facilities often utilize closed water supply systems equipped with a variety of sensors for monitoring water parameters and integrated visual control systems. This allows the creation of optimal conditions for the life and development of hydrobionts, ensuring effective management of the internal environment. Visual control using cyber-physical systems provides real-time monitoring of pool processes as well as tracking the condition of fish. This paper proposes a method using a neural network algorithm to determine the activity of fish movement under production conditions.