DEEPSPILLNET: A REAL-TIME AI FRAMEWORK FOR ENHANCED OIL SPILL DETECTION IN MARINE ENVIRONMENT USING YOLO-MOBILENET FUSION

Science and Tech

This research presents DeepSpillNet, an advanced deep learning framework that combines YOLO-V3 and MobileNet architectures for real-time, high-accuracy oil spill detection in marine ecosystems. Addressing the limitations of conventional methods—such as slow processing, high false-alarm rates, and poor adaptability to dynamic ocean conditions—the proposed model leverages MobileNet’s lightweight feature extraction and YOLO-V3’s efficient object detection to achieve superior performance. Trained on Sentinel-1 SAR imagery, the system attains 94% precision and 91% recall in spill identification, with an inference time of just 45ms, enabling rapid response to environmental hazards. Comparative analysis demonstrates a 15% improvement in F1-score over existing CNN and SAR-based approaches, while maintaining computational efficiency for scalable deployment. The framework’s robustness across diverse spill sizes and environmental conditions makes it a practical tool for marine conservation agencies, aligning with global sustainability goals. Future work will explore multi-modal data integration and edge-computing optimization to further enhance real-world applicability.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.