Japanese scientists develop AI to detect seabed plastic waste

DeepLitterAI development

Japan Agency for Marine-Earth Science and Technology researchers have developed an artificial intelligence system capable of detecting plastic waste on the seabed at processing speeds nearly double those of human analysts, according to a study published Thursday. The system, named DeepLitterAI, was trained on approximately 12,000 images collected from seabed footage captured by the agency since 1983, featuring both small litter objects and non-litter items such as rocks and marine creatures that often complicate identification efforts.

Image recognition training

Researchers utilized various pattern-recognition techniques including blurring and image inversion to minimize false detections, enabling the system to identify litter types and quantities even when objects appeared as small as five to ten percent of the image width during testing. DeepLitterAI successfully identified 80 percent of major litter items including plastic bottles and polythene bags with a margin of error of roughly 10 percent compared to expert visual inspections, the study reported.

Real-time monitoring applications

"We can quickly identify areas where large amounts of waste accumulate and use this information to implement countermeasures," said biological oceanographer Ryota Nakajima, a member of the research team. The development comes as environmental scientists increasingly focus on seabed pollution, given that much of the plastic waste entering oceans eventually sinks below the surface where detection remains technically challenging. Real-time monitoring capabilities are expected to guide targeted cleanup operations in identified high-accumulation zones.