AI on Watch: A Russian System to Detect Marine Litter in the Arctic
Russian scientists are deploying artificial intelligence to monitor marine pollution in the Arctic, offering a practical response to one of the most pressing environmental threats facing the region.
An Ecological Hotspot Under Pressure
The Arctic is more than a land of ice and extreme weather. It is one of the planet’s most fragile ecological zones and a future artery of global logistics through the Northern Sea Route. At the same time, its ecosystems are increasingly affected by human activity, including marine pollution that is difficult to detect and even harder to remove.
According to the United Nations, about 8 million tons of marine litter enter the world’s oceans every year. In the Arctic, less than one percent of that volume is estimated to be present—roughly 1,200 tons. Even these comparatively small amounts have a disproportionate impact. In cold waters, waste does not decompose for decades, becomes trapped in ice, accelerates ice degradation, and threatens marine life.
Monitoring pollution in the Arctic is particularly challenging. The region’s vast scale, severe weather, glare from ice and water, and constant vessel movement make traditional observation methods inefficient and expensive.
Teaching AI to Spot Marine Litter
To address this challenge, scientists from the Moscow Institute of Physics and Technology (MIPT) and the Shirshov Institute of Oceanology of the Russian Academy of Sciences have developed an AI-based system capable of automatically detecting marine debris on the surface of Arctic waters.
Installed on ships, the system analyzes video streams from onboard cameras in real time, identifying foreign objects among waves, reflections, and spray. The neural network was trained using data collected during a scientific expedition in 2023, with researchers processing more than 500,000 images of the Barents and Kara Seas.
The model is trained to recognize four categories of objects: marine litter, birds, sunlight glare on water, and water droplets on camera lenses. Its key strength lies in its resilience to harsh marine conditions and its ability to detect relatively small objects—critical for early identification of pollution before debris freezes into ice or sinks.
The reported detection accuracy reaches 0.4, significantly outperforming commonly used algorithms such as YOLO, which achieved accuracy levels of around 0.1 for this task. This improvement makes continuous, large-scale monitoring feasible in environments where human observation is limited.
From Arctic Protection to Global Relevance
Artificial intelligence is already playing a growing role in Russia’s Arctic environmental monitoring efforts. For example, a neural network developed by Yandex is used to identify six types of waste on aerial images of coastlines, supporting cleanup operations. Other AI systems created by MIPT and the Institute of Oceanology are used to predict extreme weather events in the Arctic more quickly and accurately than global forecasting models—an essential capability for safe navigation and regional development.
The new marine litter detection system could be integrated into the infrastructure of the Northern Sea Route to provide environmental oversight of shipping activity. Its algorithms could also be adapted for use with autonomous drones, satellites, or automated waste-collection platforms.
Beyond the Arctic, the technology has clear export potential. Marine pollution is a global problem, and an AI solution proven in one of the world’s harshest environments could be valuable for international environmental organizations working to protect oceans worldwide.
Rather than creating another experimental AI model, Russian researchers have delivered a field-ready tool with direct environmental impact—demonstrating how artificial intelligence can support sustainability goals in regions where ecological damage is both hard to reverse and easy to overlook.
