Waste Object Detection Using Mini Submarine in Water Bodies Using Adaptive Multi-Modal Fusion Transformer

In the modern era, water pollution poses a serious threat to marine ecosystems, leading to damage to aquatic life and in turn, a disastrous impact on long-term environmental threat. The system proposes advanced computer vision techniques to detect and classify the waste trash in the underwater environment. Methods like You Only Look Once (YOLO) and AUVs(Autonomous Underwater Vehicles) have practical limitations that reduce their effectiveness in underwater settings. To overcome the challenges and constraints, a novel algorithm Adaptive Multi-Modal Fusion Transformer (AMFT) is proposed for waste detection in an enhanced way. The proposed work compares the contemporary object detection framework, YOLO model, and AMFT. The system uses datasets with images of underwater with trash including plastic wastes, aluminum wastes and other debris for training and testing. The advantage of the proposed work is cost effectiveness and scalability. Waste hotspots can be identified and cleanup operations can be performed with high priority. The experimental results infer that AMFT based detection system performs extraordinarily well by achieving accurate waste recognition in different underwater conditions, like varying light levels and turbidity. The combination of advanced computer Vision techniques and real-time detection capabilities, the system offers a sustainable and scalable solution for identifying underwater waste, for cleaner water bodies and a more sustainable future. The experimental results showed that the accuracy of the clear water system exceeded 94%, while in dirty water conditions, accuracy decreased slightly but remained above 82 %