YOLO-Aqua: Intelligent Detection of Plastic Debris in Aquatic Environments

The use of plastic in water bodies has become a major global environmental problem, threatening biodiversity and disrupting food chains in the oceans. Conventional methods of detecting plastic waste in water are often ineffective, as they rely heavily on manual input, especially in shifting and complex underwater environments. This project proposes an automated and intelligent system for detecting underwater plastic. It integrates the state-of-the-art object detection model YOLOv8 with Particle Swarm Optimization (PSO). YOLOv8 enables real-time detection with a lightweight, anchor-free architecture, making it effective in identifying irregularly shaped and partially concealed plastic debris in underwater video streams. However, its performance declines when hyperparameters are manually tuned. To address this, PSO optimizes critical parameters such as learning rate, confidence threshold, and anchor box sizes, thereby improving both detection accuracy and model robustness. The system leverages PSO to extract relevant features for efficient training, while YOLOv8 identifies plastic waste in real time. A Convolutional Neural Network (CNN) is employed to detect various types of debris, including bottles, bags, nets, and covers. This approach delivers low latency, higher classification accuracy, and reduced false positives, even under murky conditions. Experimental results on standard datasets demonstrate that this method outperforms traditional YOLO models, achieving significant improvements in accuracy, precision, recall, and F1 score. The proposed system can be deployed in real time on autonomous underwater vehicles (AUVs) and large-scale ocean surveillance platforms to track plastic and support effective ocean cleanup efforts. By combining deep learning with environmental sustainability, this project provides a scalable solution to address marine pollution while advancing global conservation goals.