Exploring Deep Learning for Underwater Plastic Debris Detection and Monitoring

Abstract

In this paper, we present a comparative evaluation of state-of-the-art deep learning models for object detection in underwater environments, with a focus on marine debris detection. We investigate the performance of four prominent object detection models: Faster R-CNN, SSD, YOLOv8, and YOLOv9, using two different datasets: TrashCAN and DeepTrash. Through quantitative analysis, we evaluate the accuracy, precision, recall, and mean average precision (mAP) of each model across different object classes and environmental conditions. Our results show that YOLOv9 consistently outperforms the other models, demonstrating superior precision, recall, and mAP values on both datasets. Furthermore, we analyze the stability and convergence behavior of the models during training, highlighting the excellent stability and adaptability of YOLOv9. Our results underscore the effectiveness of deep learning-based approaches in marine debris detection and highlight the potential of YOLOv9 as a robust solution for environmental monitoring and intervention efforts in underwater ecosystems.