A neural network helped clean Kamchatka beaches.

Specialists from the Yandex School of Data Analysis have developed a neural network that significantly increased the efficiency of coastal waste cleanup in Kamchatka. The technology was applied in Kronotsky Nature Reserve и South Kamchatka Nature Reserve.

The system, developed with support from Yandex B2B Tech and Far Eastern Federal University, analyzes aerial images and determines the quantity and types of contaminants. This allowed the researchers to obtain precise data on the waste structure: 33-39% is plastic containers and packaging, and 27-29% is industrial fishing waste.

The neural network classifies waste into six categories with over 80% accuracy, mapping its location, weight, and composition. This technology has reduced the time required to remove five tons of waste by four times, optimizing the project's resources.
The technology can be taught to find other types of waste. The development code is also available in open access, which makes it accessible to other environmental initiatives and not only in Russia.