Classification and Counting of Ships Using YOLOv5 Algorithm
Rendell Sheen S. Suliva, Clint Aldrin A. Valencia, Jocelyn F. Villaverde
Abstract
Computer vision has been aiding various industries in making work efficient. In the case of marine and ocean-related industries, climate change, greenhouse gasses, fishing exploitation, and coastal contamination are all causing significant effects on human life. In the Philippines, the same problem that burdens most coastal countries exists. The system implemented to aid this problem is limited to those with access to the SAR and has no local or small-scale implementation. Different studies focus on the utilization of algorithms to detect and classify ships in the sea. Therefore, counting and classification based on the type of ships are essential. Using the YOLOv5 and DeepSORT algorithm, the system was able to achieve a model, prototype, and counting accuracy of 98.65%, 98.11%, and 100% respectively. Some of the misclassifications are due to the close similarities of the different classes and the under representation of some classes. It can be concluded that the produced model is accurate in detecting, classifying, and counting ships based on type.