Hybrid Deep Learning Model for Detecting Underwater Naval Mines

Abstract

In the realm of maritime warfare, naval mines stand as formidable threats, strategically positioned explosives lurking beneath the water's surface, poised to inflict damage upon unsuspecting ships or submarines. This proposed model harnesses the power of deep learning methodologies to discern and identify these submerged hazards. Leveraging contemporary advancements in deep learning technology, our aim is to construct robust and economical models capable of reliably detecting naval mines. Through this study, an array of deep learning models is employed to gauge their efficacy, utilizing accuracy as a primary metric for comparison and evaluation.

Among the models employed were CNN (Convolutional Neural Network), YOLOv5, VGG-19, and a hybrid fusion of CNN-VGG-19 and CNN-MobileNet. Remarkably, CNN showcased outstanding performance, achieving an impressive accuracy rate of 98%. Additionally, YOLOv5 demonstrated robust performance, closely trailing behind with an accuracy score of 97%.Surpassing them all, the hybrid models, specifically the CNN-VGG19 and CNN-MobileNet fusion, showcased the highest accuracy, reaching an outstanding 99%.