Evaluating Computer Vision Architectures for Ship Classification: A Comparative Study

Mateus R Cruz, Felipe Augusto Pereira de Figueiredo, Samuel Mafra, Danilo Machado de Oliveira, Eduardo Henrique Teixeira

DOI: 10.14209/sbrt.2023.1570923801

Evento: XLI Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2023)

Keywords:

Abstract

The complexity and vastness of the ocean make maritime monitoring a critical aspect of both national security and international trade. This paper investigates the potential of computer vision (CV) technology for monitoring maritime activity, with a particular emphasis on ship classification. The study compares various CV architectures and techniques, highlighting their relative strengths and weaknesses. The main objective of using CV in this context is to identify and track vessels and other objects of interest, which could assist in enforcing maritime regulations, increasing trade efficiency, and detecting security risks. The paper presents numerical results obtained during the training and validation of these architectures, providing valuable insights into how CV technology can be employed to improve maritime monitoring.

Download