Fig 1.
The YOLOv8 Network Model.
Fig 2.
Water surface interference factors.
Fig 3.
Bi-level attention mechanism.
Fig 4.
BiFormerBlock, C2fBF.
Fig 5.
Comparison of trajectories of various optimization algorithms in a two-dimensional loss function space.
Fig 6.
Comparison of Sophia and Adam in terms of mAP.
Table 1.
Comparison among the tested models after applying the improved loss function.
Table 2.
Comparison among the tested models after applying the improved optimizer.
Table 3.
Results of the ablation experiment.
Fig 7.
Comparison between the mAPs of YOLOv8-SST and YOLOv8n.
Table 4.
Performance comparison among different models.
Fig 8.
mAP@0.5 Changes Relative to the YOLO Series of Algorithms.
Fig 9.
Visualization of water-surface object detection for different object sizes.
Fig 10.
Visualization of water-surface object detection under different backgrounds.
Fig 11.
Visualization comparison between YOLOv8n and YOLOv8-SST.
Fig 12.
Model performance under diverse illumination environments.