Advanced Ship Detection System using Yolo v7

Accurate ship detection is essential for a wide range of maritime applications. This research paper presents an innovative ship detection method that integrates the YOLOv7 (You Only Look Once) object detection algorithm with the Jetson Nano platform. The proposed methodology combines the power of YOLOv7’s real-time and accurate ship detection capabilities with the efficient inference capabilities of the Jetson Nano platform. The YOLOv7 model is trained and evaluated on two datasets: MASATI v2 and ShipRSImageNet. The results demonstrate impressive accuracy and computational effectiveness, with mAP values of 0.76 and 0.93 on MASATI v2 and ShipRSImageNet, respectively. Additionally, precision values of 0.8 and 0.96, and recall values of 0.77 and 0.84 are achieved on the respective datasets. The integration of the trained model with the Jetson Nano platform enables efficient ship detection in real-time on live video streams. This research emphasizes the significance of appropriate model architecture and training strategies, showcasing the potential of the proposed method for accurate ship detection in dynamic marine environments and high-resolution remote sensing images. The findings of this study contribute to advancements in ship detection technology and its applicability in various maritime domains.