An Effective Object Detection Model to Detect Floating Macroplastic Debris on the Waterways Using YOLO

The escalating problem of plastic pollution in aquatic environments poses a significant threat to global ecosystems. Among the various forms of plastic debris, floating macroplastics are particularly persistent and visually conspicuous, demanding effective monitoring and quantification strategies. Traditional methods for surveying plastic pollution, such as manual collection and visual observation, are often laborintensive, geographically limited, and struggle to provide the comprehensive data required for effective mitigation strategies. However, existing detection methods primarily rely on manual surveys or simplistic computer vision approaches, which can be inaccurate and inefficient. To enhance the detection efficiency of macroplastics in waterways, a deep learning-based model was utilized, incorporating CSPDarknet, Bidirectional Feature Pyramid Network (BiFPN), and YOLOv7. This comprehensive approach addresses the global challenges introduced by macroplastic pollution and assesses its implications for environmental sustainability and public health. The experimental findings reveal that the proposed model showcases remarkable effectiveness and impressive efficiency, enabling the seamless automated identification of macroplastics on the waterways. The model delivers impressive detection accuracy (mAP) of 93.2 percent and effectively adapts to variations in object size and image quality. The model generates remarkable and more effective results when compared with the existing methods. Furthermore, its accuracy has improved by 2.3 to 2.9 percent.