An AI-Powered System for Real-Time Underwater Plastic Detection and Classification Using YOLOv7 and VGG16

Abstract

Plastic pollution in marine environments poses a severe threat to aquatic ecosystems and global biodiversity. The traditional methods used to monitor the wastes in the sea are time consuming and in virtually every situation, not real time responsive. In order to achieve effective and efficient plastic underwater detection and classification, the current paper proposes a hybrid architecture that incorporates YOLOv7, a newly emerging object detector architecture, and VGG16, a deep convolutional neural network. The system is able to detect the plastic waste and classify it into different categories, such as bottles, bags, nets, and wrappers. An SMTP notification system is proposed to simplify the detection, while an email notification with annotated detection result is recommended to be sent at each plastic detection. It could be to train YOLOv7 on large amounts of underwater images to predict where they are in general and then use VGG16 to predict exactly where they are in cropped regions. The tests show that the system is highly accurate and performant under extreme conditions in the under-sea environment, including turbidity or low visibility. The automatic alerter also provides early alerting and immediate response and remediation. Overall, the book is a brilliant, measured, terrestrial effort to solve the problem of marine waste disposal and contributes to the body of literature on environmental protection and building a sustainable ecosystem.