251129 EP丨Underwater Object Detection Dataset with Complex Scene (CSUOD)

The CSUOD dataset collects 1135 real underwater scene samples and performs manual annotation and resolution standardization, covering a variety of aquatic organisms and including different water environments (clear/turbid), deep water layers (diving layer/deep water layer), and lighting conditions (natural light/artificial light source). These images truly reflect the coupling effect of multiple environmental factors such as color cast, haze effect, and non-uniform lighting. The CSUOD dataset contains a total of 2147 annotated objects, with category distributions covering fish (46.1%), divers (16.0%), jellyfish (14.3%), turtles (10.9%), shrimp (6.9%), and squid (5.8%). The CSUOD dataset can be used for robust training and performance evaluation of underwater object detection models in complex scenarios. To quote this data, you must quote the following original papers:Citation:HOU Guojia, MA Jiaqi, WANG Yuechuan, HUANG Baoxiang, LI Kunqian. UWF-YOLO: A Lightweight Framework for Underwater Object Detection via Redundant Information Optimization[J]. Journal of Electronics & Information Technology. doi: 10.11999/JEIT251129