Marine pollution, particularly from plastics, is a pressing environmental issue, with over 11 million tons of plastic waste entering the oceans each year. A joint initiative by Yandex B2B Tech, Yandex School of Data Analysis, and the Far Eastern Federal University has led to the development of an open-source neural network designed to improve the efficiency of coastal waste cleanup in remote areas. This technology, already tested in the South Kamchatka Federal Nature Reserve, is now being expanded to the Arctic and other challenging environments.
The neural network automates the detection and analysis of waste, significantly reducing the time and resources required for pollution assessment. In Kamchatka, it enabled volunteer teams to remove 5 tons of waste four times faster than traditional methods, identifying plastic containers, packaging, and industrial fishing debris as major components of coastal pollution. With an accuracy rate of over 80% in waste detection, the tool leverages computer vision and semantic image segmentation to map and remove pollutants efficiently.
This innovation not only addresses the logistical challenges of cleaning up remote coastal areas but also aligns with global efforts to combat plastic pollution, as highlighted by the upcoming World Environment Day 2025. The open-source nature of the neural network ensures its adaptability for various environmental monitoring and cleanup tasks worldwide, marking a significant step forward in the use of artificial intelligence for environmental conservation.
