Japanese researchers unveil AI tool to map plastic debris on ocean floor

Scientists at the Japan Agency for Marine-Earth Science and Technology have introduced DeepLitterAI, an AI platform designed to identify plastic litter on the seafloor. They compiled a collection of about 12,000 images from footage taken since 1983, deliberately including both waste and non-waste objects such as rocks and sea creatures to teach the algorithm to avoid misclassification. Training incorporated image-alteration methods like blurring and inversion, which lowered erroneous detections.

In field trials, the system recognized the type and amount of debris even when it covered only a small portion of the image, correctly spotting roughly 80% of items like bottles and bags while maintaining a margin of error near 10% compared with expert visual checks. Processing speed was nearly double that of human analysts, turning a task that would normally require a month into a matter of days, and overall accuracy improved by 1.6 times. Researchers say the technology could enable rapid identification of pollution hotspots and support timely mitigation measures.