Enhancing ROV Navigation with Object detection using Yolov8 and Conv2d layer

Detecting objects underwater is no easy feat, hindered by the conditions and intricate backgrounds of these submerged environments. While YOLOv8 has made strides in object detection performance across various datasets, its design is primarily optimized for land-based detection, leaving room for improvement in aquatic scenes. To tackle this limitation, we present a groundbreaking approach that combines the power of YOLOv8 with Conv2D layers. Our solution involves enhancing the YOLOv8 architecture specifically for underwater object detection by incorporating a Conv2D layer into its backbone network. This enables the extraction of more resilient features from underwater images. Moreover, we introduce a novel loss function that addresses the unique challenges of detecting objects in these challenging conditions, such as diminished visibility. Adding conv2d layer to the yolov8 model will in crease the efficiency and accuracy of the object detection.