This book focuses on a comprehensive investigation into data-driven Sea State Estimation (SSE) by leveraging a vessel’s own motion data. It presents a collection of advanced deep learning frameworks designed to overcome critical, real-world challenges inherent in this approach. This book systematically introduces key issues including: the class imbalance of sea state data, where rare but hazardous conditions are difficult to predict; the need for model transferability between different ships and loading conditions; and the crucial demand for security and robustness against adversarial data attacks. To solve these problems, the book introduces a suite of innovative architectures employing techniques such as densely connected convolutional networks, prototype-based classifiers, multi-scale feature learning, adversarial transfer learning, and dynamic graph networks. The efficacy of these models is rigorously validated on both public benchmarks and specialized ship motion datasets, demonstrating superior performance over existing state-of-the-art methods and providing a robust toolkit for enhancing maritime safety and efficiency.

This book focuses on a comprehensive investigation into data-driven Sea State Estimation (SSE) by leveraging a vessel’s own motion data. It presents a collection of advanced deep learning frameworks designed to overcome critical, real-world challenges inherent in this approach. This book systematically introduces key issues including: the class imbalance of sea state data, where rare but hazardous conditions are difficult to predict; the need for model transferability between different ships and loading conditions; and the crucial demand for security and robustness against adversarial data attacks. To solve these problems, the book introduces a suite of innovative architectures employing techniques such as densely connected convolutional networks, prototype-based classifiers, multi-scale feature learning, adversarial transfer learning, and dynamic graph networks. The efficacy of these models is rigorously validated on both public benchmarks and specialized ship motion datasets, demonstrating superior performance over existing state-of-the-art methods and providing a robust toolkit for enhancing maritime safety and efficiency.

This book focuses on a comprehensive investigation into data-driven Sea State Estimation (SSE) by leveraging a vessel’s own motion data. It presents a collection of advanced deep learning frameworks designed to overcome critical, real-world challenges inherent in this approach. This book systematically introduces key issues including: the class imbalance of sea state data, where rare but hazardous conditions are difficult to predict; the need for model transferability between different ships and loading conditions; and the crucial demand for security and robustness against adversarial data attacks. To solve these problems, the book introduces a suite of innovative architectures employing techniques such as densely connected convolutional networks, prototype-based classifiers, multi-scale feature learning, adversarial transfer learning, and dynamic graph networks. The efficacy of these models is rigorously validated on both public benchmarks and specialized ship motion datasets, demonstrating superior performance over existing state-of-the-art methods and providing a robust toolkit for enhancing maritime safety and efficiency.