TOKYO - A Japanese team of researchers has developed an artificial intelligence system to better detect marine plastic waste on the seabed, it said in an international journal on Thursday.
A team from the Japan Agency for Marine-Earth Science and Technology has developed a system called DeepLitterAI. Capable of processing data at almost twice the speed of humans, it is expected to be utilized for real-time monitoring of marine plastic waste.
The team created a dataset of approximately 12,000 images based on footage of Japan's seabed captured by the agency since 1983, featuring small litter objects, as well as non-litter objects such as rocks and creatures that are easily misidentified, according to the paper in Environmental Pollution.
The team then trained the system using various patterns, such as blurring and inversion, to reduce false detections.
While much of the plastic waste that flows into the sea sinks to the seabed, it is said to be difficult to grasp the situation accurately.
Ryota Nakajima, a biological oceanographer on the team, said, "We can quickly identify areas where large amounts of waste accumulate and use this information to implement countermeasures."
During tests using actual footage, the system identified the type and quantity of litter even when it appeared as small as 5 to 10 percent of the image width, while successfully detecting 80 percent of major litter items, such as plastic bottles and polythene bags.
The margin of error compared to visual inspection by experts was around 10 percent, and the system can complete an analysis that would take a human about one month in just a few days.
Given deep sea footage is captured with a wide-angle lens, existing AI systems would only learn to recognize large items of litter, often leaving smaller ones undetected. The team reports that this new method has improved accuracy by 1.6 times.
