Fig 1.
Corrosion of metallic facilities in the marine environment.
Table 1.
Advantages and disadvantages of different corrosion detection methods.
Fig 2.
YOLOv5s model structure framework.
Fig 3.
CBG-YOLOv5s model structure framework.
Fig 4.
CBAM model structure framework.
Fig 5.
BiFPN-CBAM structure framework.
Fig 6.
(a) Ordinary convolution operation. (b) Ghost convolution operation.
Fig 7.
C3Ghost module.
Fig 8.
Processing flow of metal surface corrosion data.
Fig 9.
Corrosion of different types of metal surfaces.
Fig 10.
Demonstration of the effects of data enhancement.
Table 2.
Division of metal surface corrosion datasets.
Table 3.
Experiment environment.
Table 4.
YOLOv5s model performance.
Table 5.
CBG-YOLOv5s model performance.
Table 6.
Performance comparison between YOLOv5s and CBG-YOLOv5s.
Fig 11.
YOLOv5s detection effect diagram.
Fig 12.
CBG-YOLOv5s detection effect diagram.
Table 7.
Comparison with typical models.
Fig 13.
Recognition effect of each model.
Table 8.
Ablation experiments.
Fig 14.
Curve of changes in each evaluation indicator during the training process.