@article{manjur_skarke_gurbuz_2025, title={Detection and Classification of Gas Seeps in MBES Imagery Using CFAR and Feature-Based Learning}, DOI={10.23919/oceans59106.2025.11245025}, author={Manjur, Sultan Mohammad and Skarke, Adam and Gurbuz, Ali Cafer}, year={2025}, month={Sep} }
 @article{manjur_skarke_gurbuz_2025, title={Evaluating Depth-Dependent Variability in Machine Learning-Based Seafloor Gas Seep Detection Using Multibeam Echosounder}, DOI={10.23919/oceans59106.2025.11245000}, author={Manjur, Sultan Mohammad and Skarke, Adam and Gurbuz, Ali Cafer}, year={2025}, month={Sep} }
 @article{manjur_senyurek_kalski_skarke_gurbuz_2025, title={Machine learning-based seafloor gas seep detection in sonar water column images}, volume={13482}, DOI={10.1117/12.3054109}, abstractNote={Seafloor gas seep detection is a critical area of research with implications for ocean carbon cycling, chemosynthetic ecology, energy exploration, and geohazard monitoring. Multibeam echosounders, which utilize acoustic impedance differences to generate detailed acoustic imagery, including bathymetry and water column features, offer a means to map the seafloor with high precision. However, detecting gas seeps traditionally relies on human interpretation of sonar water column imagery, which is time and labor intensive, costly, and prone to inconsistency due to annotator experience and variability. While in-situ techniques provide detailed, localized insights, they are inefficient for large-scale applications. This creates a significant gap in the ability to efficiently and accurately identify seafloor gas seeps in a cost-effective and scalable manner. Computer vision and machine learning techniques offer a transformative approach to efficiently process large-scale sonar data with high accuracy. This study proposes a novel convolutional neural network (CNN) framework for seafloor gas seep detection, leveraging its ability to extract complex spatial features from water column images. To ensure robustness, we implemented location-based cross-validation, a rigorous validation strategy that mitigates spatial bias and better reflects real-world scenarios. Furthermore, we employed explainable AI techniques to gain insight on the regions of the water column data that the model focused on during decision making, enabling seep localization and enhancing interpretability. Our proposed approach achieved an impressive accuracy of 88% with 77% precision and 72% recall, showcasing a significant leap toward automated, explainable, and scalable solutions for seafloor gas seep detection with extraordinary potential for real-world applications.}, journal={OCEAN SENSING AND MONITORING XVII}, author={Manjur, Sultan M. and Senyurek, Volkan and Kalski, Ramon and Skarke, Adam and Gurbuz, Ali Cafer}, year={2025}, month={May} }