Underwater Object Detection
Overview
Underwater object detection studies how to locate and classify fish, debris, divers, and other targets in images affected by color loss, blur, backscatter, clutter, or sonar noise. Common approaches adapt YOLO-style detectors with multi-scale feature pyramids, attention, and physics-aware image restoration; sonar work also explores spiking networks, 3D detection, and fusion with cameras. Evaluation spans optical datasets such as RUOD, DUO, and URPC and sonar datasets such as UATD, with newer benchmarks examining domain differences, annotation quality, and cross-modal robustness. A central challenge is balancing accuracy, small-target sensitivity, and real-time efficiency while ensuring models generalize across changing water, seabed, sensors, and acquisition conditions.
Latest papers
29 papers2 papers in the last four weeks, down 60% on the four weeks before. 0.0% of all new papers.