0
Article ? AI-assigned paper type based on the abstract. Classification may not be perfect — flag errors using the feedback button. Tier 2 ? Original research — experimental, observational, or case-control study. Direct primary evidence. Sign in to save

Marine Debris Detection Using Artificial Intelligence

Original title: Marine Debris Recognition with Artificial Intelligence (Detecção de resíduos marinhos utilizando inteligência artificial)

Zenodo (CERN European Organization for Nuclear Research) 2026
João A. B. Conceição, Victor H. M. Nascimento, Leonardo T. Aguiar, Guilherme A. A. Barros, Marise Miranda

Summary

Every year, about 11 million tons of plastic pours into our oceans, breaking down into tiny microplastics that work their way up the food chain and onto our plates. Researchers built an AI tool that scans underwater photos and automatically spots marine trash with about 82.5% accuracy, then packaged it into a free web platform so cleanup groups and governments can find and target pollution hotspots faster. Catching this waste before it fragments further could help slow the flow of microplastics into seafood and, ultimately, into us.

Body Systems

Abstract Marine pollution, particularly solid waste and plastics, poses an escalating threat to marine ecosystems, biodiversity, and human health. Contrary to the outdated perception of oceans as limitless waste repositories, it is now understood that plastic materials persist for centuries, fragmenting into microplastics that infiltrate the food chain. Research indicates that approximately 11 million metric tons of plastic enter the oceans annually, contributing to an accumulated total of nearly 200 million tons. In Brazil, this issue is exacerbated by intense port activities and insufficient waste management infrastructure. This reality underscores the urgent need for effective monitoring and prevention strategies. This study proposes a machine learning-based solution for the automatic detection and classification of marine debris. We evaluated multiple architectures, including K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Convolutional Neural Networks (CNN), to determine the most effective approach for identifying pollution in underwater imagery. The results demonstrate that the CNN model achieved the highest performance, with 82.5% accuracy. Furthermore, we developed a web-based platform to democratize access to these environmental insights, aiming to assist NGOs and government agencies in mitigating critical pollution hotspots.

Share this paper