YOLOv8-Based Detection of Plastic Pollution in Underwater Environments

Underwater plastic pollution poses a severe threat to marine ecosystems and human health, necessitating advanced solutions for its detection and mitigation. This study leverages the YOLOv8 model, a state-of-the-art deep learning framework, to address the complexities of underwater plastic detection. Utilizing a robust dataset of underwater plastic images, the model was fine-tuned through preprocessing techniques, such as image enhancement and augmentation, to overcome challenges like varying lighting conditions, water turbidity, and occlusions. The training process demonstrated a consistent reduction in loss and significant improvements in accuracy, with the final training and validation accuracies reaching 99.95% and 98.56%, respectively, over 30 epochs. Similarly, the training and validation losses converged to minimal values, indicating the model’s strong generalization capabilities. The results highlight the model’s ability to effectively identify diverse plastic objects under challenging underwater conditions, making it a promising tool for largescale marine monitoring and environmental conservation. This research not only showcases the efficiency of YOLOv8 in real-time underwater plastic detection but also sets the foundation for integrating deep learning with autonomous marine systems for scalable pollution monitoring and cleanup efforts. Future work will focus on deploying the model in real-world scenarios and enhancing its robustness against more diverse environmental conditions.