The Yandex neural network helped volunteers find and clean hard-to-reach coasts from garbage four times faster

2025: Use of Yandex neural network in the Clean Coast project

Yandex, FEFU and the Conservationists Foundation shared the results of the Clean Coast project to organize the cleaning of the sea coast using neural networks. Since the start of the project in 2024, volunteers have cleared more than 50 km of hard-to-reach protected coasts in Kamchatka, the Leningrad Region and the Primorsky Territory of garbage. In 2026, the neural network will be used in six more protected natural areas in different parts of the country, from Kaliningrad to Kamchatka, for system cleaning and monitoring of coastal conditions, as well as in the Arctic. Analyzing this data will help environmental scientists develop new strategies to combat pollution. Yandex announced this on December 9, 2025.

The solution was developed by experts from the Center for Technology for Yandex Society and ML developers of the School of Data Analysis with the support of FEFU scientists. The neural network analyzes aerial photographs of coasts and classifies waste into six types - fishing nets, metal, rubber, large plastic, concrete and wood - with an accuracy of more than 80%. The model marks the coordinates on Yandex Maps, determines the composition and weight of garbage, which allows volunteers and environmental control services to assess the amount of work and form teams for cleaning territories. During the expeditions in 2025, the neural network was trained on another 20 thousand aerial photographs.

{{quote'In 2025, the geography of the project has expanded significantly - new reserves and national parks have joined the initiative, and in 2026 we will be able to move from one-time cleaning to a full-fledged system cleaning of the coasts. Neural networks help us in this: they allow us to accurately assess the scale of pollution, plan work and allocate resources more efficiently - and work 4 times faster than without the use of technology. We see how neural networks make environmental projects more effective and predictable, which means they help to really improve the state of the territories, - said Roman Korchigin, project curator and project director of the Conservationists Foundation. }}

The problem of pollution of the Arctic coast is a consequence of global environmental processes. Atlantic currents nail garbage to the beaches of Franz Josef Land and Novaya Zemlya, where the degradation of plastic waste freezes and the rate of waste accumulation grows. It is important for scientists to investigate the routes of movement of garbage and conduct exploration of bottom sediments of plastic, since the future of the Arctic and the North and the species that live there depend on this. The use of neural networks is the key to the work of researchers in harsh Arctic conditions, where work is difficult and resources are limited and require careful planning, "said Artyom Smolokurov, head of the Clean North - Clean Country movement.

In 2025, work was carried out on three federal specially protected natural areas: in the Kronotsky Reserve and the South Kamchatsky Reserve, the Lower Svir Reserve on Ladoga and the Far Eastern Marine Reserve on Popova Island. In 2026, the initiative will be joined by the national parks "Curonian Spit," "Land of the Leopard," "Commander Islands," Arctic territories and the Dagestan Reserve.