DeepLitterAI: Automated detection and quantification of deep-sea benthic plastic and macrolitter with field validation in waters around Japan.

DeepLitterAI achieves up to 0.81 average precision in detecting deep-sea benthic plastic and macrolitter, outperforming large-object models by 1.6 times in challenging conditions.

• Why it matters: Monitoring deep-sea litter is crucial for understanding pollution impacts, but manual inspection is slow and limited in scope, hindering large-scale assessments and standardized monitoring efforts.
• What they did: Developed using 12,029 annotated images from the JAMSTEC dataset, DeepLitterAI combines YOLOv11x and BoT-SORT to automatically detect and quantify small litter objects in deep-sea imagery, validated across multiple depths and conditions.
• The result: The system enables faster, scalable, and more consistent deep-sea litter monitoring, with AI analysis being 2.1 times quicker than manual methods and providing robust performance despite viewing angle variations.