Marine ship detection and classification based on YOLOv5 model. Issue 1 (1st January 2022)

• Record Type: Journal Article
• Authors: Zhang, Xuan; Yan, Mingzhong; Zhu, Daqi; Guan, Yang
• Is Part Of: Journal of physics. Volume 2181:Issue 1(2022)
• Publication Date: 2022-01-01
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/2181/1/012025
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: An improved deep learning neural model YOLOv5-DN based on YOLOv5 is proposed for marine ship detection and classification in the area of harbours and heavy traffic waterways. The CSP-DarkNet module in YOLOv5 is replaced by CSP-DenseNet to promote the accuracy of target detection and classification in the proposed model. Sample marine ships in the data set are divided into six classes: ore carriers, general cargo ships, bulk cargo ships, container ships, passenger ships, and fishing ships to meet the detection needs in the areas of ports and waterways. The data set are grouped into a training set, testing set, and validating set by the proportion of 6:2:2. Experiments show that the improved model has better average accuracy, from 62.2% to 71.6%.