Ocean Vue Acoustic Analysis in Oceanography

Abstract

In today’s maritime operations, there exists a challenge—how to accurately and efficiently classify ships based on their acoustic signatures. This problem extends across various domains, including security, environmental monitoring, and marine research. To address this challenge, we have developed a web application for ship sound classification, making use of deep learning techniques. This project represents a significant advancement in the field of ship classification, using the Resnet50 deep learning model which was trained on mel spectrograms of ship acoustic signatures. The model’s performance is proof of the project’s dedication to achieving high-precision results. With an impressive validation accuracy of 96 this approach improves on previous benchmarks. The developed web application not only meets the objective of ship classification but also integrates a range of essential features. These include user management functionalities such as registration, authentication, and profile management. Users can easily upload ship sound samples for classification and access their classification history. The system generates reports with graphical representations of classification results, enhancing the user’s analytical capabilities. This project fills a critical gap in ship sound classification, offering an advanced solution that resonates across industries. The journey from problem identification to solution development has been marked by experimentation and fine-tuning, closing in the selection of the Resnet50 model as the main DL model for classification.