Underwater forensic analysis for target recognition faces significant challenges due to low visibility, color distortion, and degraded image quality. Detecting small targets and managing target aggregation further complicate the process, often leading to suboptimal detection results. Traditional methods struggle with large model sizes, slow detection speeds, and limited range, particularly when reliant on wired systems. With the growing need for marine exploration, efficient underwater target detection has become a critical focus in robotics and marine forensic investigations. This chapter presents an AI-driven approach using the YOLOv5 convolutional neural network for real-time underwater target recognition. Deployed on the Jetson Nano platform, this solution is trained on an enhanced underwater dataset to improve accuracy and reliability in classifying submerged objects. By leveraging the Jetson Nano, the system eliminates the limitations of traditional wired detection methods while enhancing real-time detection capabilities. Performance evaluations demonstrate that this approach effectively identifies aquatic objects even in visually degraded environments. This AI-powered solution offers a robust, scalable method for underwater forensic analysis and marine exploration, ensuring high-precision detection in challenging conditions.