Real-Time and Energy-Efficient Ship Detection Using YOLOv8 on FPGA Platforms for Marinetime Surveillance

This paper presents the implementation of a real-time and energy-efficient ship detection system using the YOLOv8 deep learning model on FPGA platforms. By leveraging the parallel processing capabilities of FPGAs, the system achieves high-speed and low-power inference while maintaining detection accuracy, addressing the limitations of traditional CPU and GPU-based solutions. Optimization techniques such as quantization and pruning adapt the YOLOv8 model to FPGA resource constraints, ensuring efficient deployment. A diverse maritime dataset, annotated with various ship types and environmental conditions, is used to train and validate the system, enabling robust performance across varying scenarios. The deliverables include a functional hardware prototype integrating a camera module for live video feed, FPGA design files for deploying YOLOv8, and detailed documentation of the system architecture, optimizations, and testing results. This research demonstrates the potential of FPGA-based deep learning for maritime applications, offering a scalable solution for port surveillance, navigation safety, and maritime security.