A Novel Underwater Marine Dataset with Diverse Scenarios for Robust Object Detection

Underwater object detection is a critical aspect of marine exploration, environmental monitoring, and underwater robotics, demanding robust and efficient solutions. This study introduces a novel dataset tailored for evaluating deep learning methods in the context of underwater object detection. We systematically compare the performance of four state-of-the-art deep learning methods. The dataset encompasses diverse underwater scenarios and challenges, ensuring a comprehensive evaluation of the methods. Our analysis provides in-depth insights into the strengths and limitations of each deep learning approach, offering guidance for selecting optimal methods across varied underwater environments. This research significantly contributes to the advancement of underwater object detection, emphasizing the role of deep learning methodologies and the importance of domain-specific datasets in benchmarking and enhancing existing methods. The experiments show that the DG-YOLO with DarkNet-53 as the backbone outperforms the other algorithms for underwater object detection in the novel tailored dataset.