DUO.zip

Aiming at the problem of low underwater recognition accuracy caused by dense and fuzzy targets in underwater target detection, an underwater target detection algorithm combining attention mechanism and downsampling is proposed.For the experimental evaluation, the DetectingUnderwater Objects (DUO) dataset was used,which was collected and re-annotated based onthe corresponding dataset. The dataset containsa total of 7782 images, including 6671 images fortraining and 1111 images for testing. The totalnumber of targets in the dataset is 74,515, ofwhich the number of holothurians, echinus, scal?lops, and starfish are 7,887, 50,156, 1,924, and14,548, respectively.