[Retracted] A Novel Attention-Based Lightweight Network for Multiscale Object Detection in Underwater Images

Table 3

Comparison of AP values of different methods for objects of different sizes. The best results are marked in bold.

Method
Yolov5	40.07	68.31	75.38	62.34	43.64
RON	43.57	72.45	78.03	65.17	45.49
RefineDet	44.21	73.49	79.07	65.42	46.03
STDN	43.61	74.08	80.15	66.39	46.55
SWIPENet	45.82	75.63	81.43	68.21	48.08
Faster R-CNN-AON	46.23	77.26	81.59	69.00	49.20
RFBNet	42.48	72.52	77.24	64.38	44.61
Ours	48.73	76.90	83.41	69.84	49.94