Automatic Determination of Sturgeon Size Using Deep Learning Technologies

Fish resources play a crucial role in Russia's economy, especially given its extensive coastline, vast water areas, and rich marine and freshwater resources. Addressing the decline and restoration of fish populations, a consequence of farming practices, illegal fishing, and environmental catastrophes, stands as a critical issue in today's world. In Russia in recent years, there has been active construction and development of fish farms, fisheries and biological laboratories, partly due to economic difficulties. One of the most valuable fish species is sturgeon, which requires special conditions of confinement compared to, for example, catfish.  Production workers monitor the living conditions of the fish, including water parameters, and observe the growth and activity of these fish. This paper analyzes the different existing methods of weighing fish in production facilities. Nowadays, there is a trend toward digitization of production and implementation of cyber-physical systems to improve the efficiency of production. Visual inspection, complemented by neural network analysis, is emerging as a valuable approach for addressing the challenge of fish weight estimation. The study suggests a viable technique for automatically measuring sturgeon sizes at various developmental phases using YoLo9 and highlights potential areas for future research.