Artificial Intelligence Era in Tracking Plastic Waste on the Seabed

Scientists have launched an artificial intelligence-supported system called DeepLitterAI to quickly and accurately detect plastic garbage on the seabed. The system can perform extensive data analysis in a much shorter time compared to humans.

Development of the DeepLitterAI System

The artificial intelligence system developed by researchers and named DeepLitterAI was designed to be used in the fight against plastic pollution on the seabed. The system allows for the real-time tracking of waste.

Dataset and Training Process

To train the system, a large dataset of approximately 12,000 photographs taken from the seabed between 1983 and 2025 was used. This set included small pieces of garbage as well as non-garbage objects such as rocks.

To prevent the artificial intelligence from perceiving wrong objects as garbage, the system was trained in detail using various techniques including blurring.

Performance and Detection Success

Having a data processing capacity twice as fast as humans, the system showed high success in tests conducted with real images. The type and amount of waste in small areas, accounting for 5 to 10 percent of the image width, could be determined.

While 80 percent of large garbage pieces such as plastic bottles and polyethylene bags were successfully detected, it was determined that the system's margin of error was around 10 percent.

Time Savings and Expert Opinions

Compared to manual inspections by experts, it was observed that the system could complete data analysis that takes humans about a month in just a few days.

Researcher and biological oceanographer Nakajima Ryota stated that areas with dense waste accumulation can be quickly identified and this data will be used to take precautions.