Underwater Object Detection Using Deep Learning

Abstract

Underwater object detection plays a crucial role in fields such as marine biology, defense, underwater robotics, and environmental monitoring. Conventional detection systems suffer performance degradation due to underwater environmental challenges such as light absorption, scattering, and color distortion. This paper proposes a deep learning-based solution utilizing convolutional neural networks (CNNs), specifically YOLOv5, for accurate and real-time underwater object detection. The model is trained on annotated underwater datasets and implemented using PyTorch. The system includes a preprocessing pipeline, a trained detection model, and a Python-based user interface for inference. Evaluations using precision, recall, and mean Average Precision (mAP) confirm significant performance improvements. The proposed solution is scalable for real-time deployment in autonomous underwater vehicles (AUVs), underwater drones, and research tools. This work demonstrates the feasibility and effectiveness of deep learning in underwater object detection and lays the groundwork for future improvements such as multi-object tracking and IoT integration.