Performance Investigation of Hybrid YOLO-VGG16 Based Ship Detection Framework Using SAR Images

Synthetic Aperture Radar (SAR) images are realized as encouraging data information for checking oceanic activities and its function for oil and ship recognizable proof, which is the focal point of numerous past research considers for better spatial goals. Several article discovery strategies extending from customary to deep learning approaches are proposed. Ship detection framework in deep learning technique accomplishes high execution, which benefits from a SAR free open dataset (SFOD). Nonetheless, a dominant part of them are computationally dangerous and have exactness issues. The main problem identified is when the number of images increases, performance may decrease. To overcome this, we propose a technique called Hybrid YOLO, which realizes K-Means Clustering and WordTree for object identification and image classification. Hybrid YOLO also realizes SEPD for the improvement between sea clutter and ship targets and bounding boxes for the probability update network. The proposed model is implemented using Conda, used with Tensorflow and Keras Framework utilizing the SAR Ship Dataset. The presentation of the Hybrid YOLO model is enhanced in rapports of accuracy and performance measures when compared with other existing models.