Figure 1.

Illustration of vision-based underwater object detection. The detection result is presented as a bounding box with a label, where $(x_i,y_i)$ denotes the ith object’s center coordinates and $(w_i,h_i)$ denotes the box’s width and height. Here, $(x,y)$ are the axis frames, with the origin in the image’s upper left corner (image from DUO dataset [6]).