Marine Debris Detection System Using Yolov11

Abstract

Marine debris has become a major environmental concern that threatens aquatic ecosystems and coastal sustainability. Efficient and accurate detection of marine waste is therefore essential for effective monitoring and cleanup efforts. This study proposes an enhanced object detection approach that integrates attention mechanisms with the YOLOv11 architecture to improve the identification of marine debris in complex ocean environments.The proposed framework combines the strong instance detection capability of YOLOv11 with attention modules, including Coordinate Attention and a Bottleneck Transformer. While YOLOv11 with Coordinate Attention demonstrates consistent performance across diverse environmental conditions, the Bottleneck Transformer contributes to identifying debris regions that may be overlooked during manual annotation. Although slightly less stable in some scenarios, the transformer-based approach shows improved performance in detecting larger debris objects, indicating its potential usefulness for specialized monitoring tasks.Experimental results demonstrate that incorporating attention mechanisms into YOLOv11 enhances detection capability and adaptability in maritime environments. The findings suggest that attention-based object detection models can effectively support environmental monitoring systems by addressing different operational requirements. Furthermore, the study highlights the importance of selecting detection architectures according to specific deployment conditions, as different models may provide advantages depending on the characteristics of the debris and the monitoring environment.