This project aims at detecting Ship-to-Ship (STS) transfers and ships in Synthetic Aperture Radar (SAR) images from Sentinel-1. The project focuses on the Laconian Gulf area and utilizes a fine-tuned YOLOv8 model to analyze SAR images. The model was trained on a curated dataset created with CVAT, which includes labeled SAR images of ships and STS-transfers. The dataset will be made available soon. Additionally, the repository includes a Dash dashboard for visualizing the results.
The primary objective of this project is to enhance maritime surveillance by accurately identifying STS transfers and ships in SAR images. This is achieved by matching detections with Automatic Identification System (AIS) data, providing a comprehensive view of maritime activities in the Laconian Gulf.
The dataset used in this project is a collection of labeled SAR images of ships and STS-transfers, created using CVAT. This dataset is crucial for training and validating the models used in the project. The link to the dataset will be updated shortly.
The repository includes a Dash dashboard designed to visualize the results of the STS transfer detection and ship identification. This dashboard provides an interactive way to explore the data, offering insights into the effectiveness of the detection models and the accuracy of the AIS data matching.
Coming soon
Contributions to this project are welcome. If you have suggestions for improvements or wish to contribute to the project, please open an issue or submit a pull request.
This project was developed by Mathias Mumm, Jordan Vinckevleugel and Richard Mallett as part of a portfolio project at the Data Science bootcamp Data Science Retreat (Berlin)
This project is licensed under the MIT License.