Sockeye salmon census using computer vision - VNIRO
25 June 2025

Sockeye salmon census using computer vision

Experimental research work on the automation of the assessment of the number of sockeye salmon spawners using computer vision is carried out at the fish counting fence (FCF) at the source of the Ozernaya River (Lake Kurilskoye).
 
Continuous counting of the Ozernaya sockeye salmon at the scientific observation post of the Kamchatka branch of the Russian State Research Center "All-Russian Research Institute of Fisheries and Oceanography" (VNIRO) has been carried out for more than 80 years. Until 2021, only visual counting was used, since 2021 - the collection of photos and videos of sockeye salmon that passed through the FCF checkpoints.
 
In 2024, as a result of processing the data array using machine learning of artificial neural networks, a model for detecting objects was created that can recognize spawning sockeye salmon in a static image or video recording, an algorithm for counting fish was developed.
 
The deployment of an artificial neural network for processing images transmitted from video cameras will automate the process of identifying and recording fish at the existing fish farm without the need for human intervention and will prevent possible subjective errors during the recording work.
 
VNIRO Press Service