Paper Title
BENEATH THE SURFACE: DEEP LEARNING FOR UNDERWATER OBJECT DETECTION, ENHANCEMENT AND TRACKING

Abstract
Underwater object detection poses considerable challenges. Inadequate visibility, diminished contrast, and color distortion hinder the extraction of reliable features and compromise detection accuracy. Conventional object detection models falter in degraded conditions, particularly when real-time performance is essential for practical applications. We propose a hybrid deep learning framework that integrates underwater image preprocessing with a YOLOv12–ResNet-50 architecture. The preprocessing phase rectifies visual quality concerns by augmenting contrast and diminishing noise, thereby enabling the network to acquire more significant features. Exhibit that this methodology attains enhanced detection precision and resilience in demanding underwater environments without sacrificing real-time efficacy. The framework demonstrates potential for practical applications in underwater surveillance and autonomous navigation systems.The framework shows promise for practical applications in underwater monitoring and autonomous navigation systems. Keywords - Underwater Object Detection, Yolov12, Resnet-50 ,Deeplearning ,Imageenhancement ,Real-Time Detection Autonomous Navigation.