Marine Debris Movement Prediction using Deep Learning and Satellite Imagery
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Marine debris poses a critical threat to aquatic ecosystems and human activities. Detecting floating debris across vast oceans is challenging due to dynamic marine conditions and the dispersed nature of pollution. In this study, we propose a deep learning-based approach leveraging commercial satellite imagery to detect marine debris efficiently and at scale. Our methodology utilizes Planetscope optical images and employs the SSD ResNet101 architecture integrated with Feature Pyramid Networks (FPN) for enhanced object detection performance. A customlabeled dataset of 1,370 marine debris instances was prepared and used for training. The model achieved a precision of 0.78 and a recall of 0.70 on the test set, demonstrating its effectiveness in distinguishing marine debris from ocean background clutter. The complete inference pipeline supports scalable tile-based processing and exports outputs in GeoJSON format for integration with geospatial dashboards. This work showcases the potential of deep learning for supporting real-time marine conservation efforts.
Fields of Study & Subjects
Microplastics and Plastic Pollution
Remote-Sensing Image Classification
Oil Spill Detection and Mitigation
Publication details
- Journal / source
- CRC Press eBooks
- Pages
- 75–81
- Publication date
- 31 March 2026
- Language
- EN
- Document type
- Book chapter