Abstract

Maritime operations in Nigerian pilotage districts face challenges in accurately classifying vessel types, which is critical for enhancing safety and operational efficiency. This study addresses the problem by exploring the efficacy of Convolutional Neural Networks (CNNs) for vessel type classification, focusing on two widely recognized architectures: ResNet and AlexNet. The objective is to evaluate their performance in this domain using a dataset comprising images of vessels commonly observed in Nigerian waters. The dataset underwent preprocessing and augmentation to improve model resilience. ResNet, leveraging residual learning to overcome the vanishing gradient issue, and AlexNet, known for its efficiency in real-time applications, were trained and validated on this dataset. The results revealed that ResNet achieved superior performance with an accuracy of 96%, precision of 95%, recall of 94%, and an F1-score of 93.5%. In comparison, AlexNet achieved an accuracy of 94%, precision of 94%, recall of 93%, and an F1-score of 92.5%. The difference highlights ResNet’s ability to better balance precision and recall while minimizing misclassifications. These findings validate the potential of CNN-based models to automate vessel classification, thereby reducing human error, enhancing critical response times, and supporting the development of advanced maritime traffic management systems in Nigerian pilotage districts.

Key words: Convolutional Neural Networks (CNNs), pilotage districts, ResNet, AlexNet, Accuracy, Precision