HD-YOLO: The hydroid of aurelia detector.

In recent years, jellyfish outbreaks have been frequent in near-shore waters and have become a worldwide problem. At present, the main species that form jellyfish disasters along the coast of China is aurelia. In order to control jellyfish outbreaks from the source of growth, this paper proposed a hydroid detector HD-YOLO improved from YOLOv7, and established a hydroid of aurelia dataset. Firstly, the original dataset was established by acquiring images of hydroids via the web, in the laboratory and undersea photography. Secondly, MSRCR, a multi-scale enhancement algorithm with color recovery, was used to improve the underwater image quality. Then the dataset is augmented by the combination of Mosaic and Mixup. Finally, the effectiveness of HD-YOLO, underwater image enhancement and dataset augmentation methods were verified through a series of comparative experiments. This study provides a more accurate and faster method for the hydroid of aurelia detection and a more generalised dataset.