Fig 1.
The structure of the EFCRFNet.
Fig 2.
The structure of the ECRF module.
Fig 3.
Illustration of Multi-Scale Edge Feature Extraction in EFEM.
Fig 4.
The structure of the EFEM module.
Fig 5.
The structure of the SA module.
Fig 6.
The structure of the CA module.
Table 1.
Evaluation indicator data analysis.
Fig 7.
The qualitative comparison between our proposed method and 16 SOTA methods, where the red dashed line represents the graph predicted by EFCRFNet.
Fig 8.
P-R curve of EFCRFNet and other 17 SOTA models.
Fig 9.
P-R curve of EFCRFNet and other 17 SOTA models.
Fig 10.
P-R curve of EFCRFNet and other 17 SOTA models.
Fig 11.
F-measure curve of EFCRFNet and other 17 SOTA models.
Fig 12.
F-measure curve of EFCRFNet and other 17 SOTA models.
Fig 13.
F-measure curve of EFCRFNet and other 17 SOTA models.
Table 2.
“No ECRF “refers to the experimental results obtained after removing the ECRF module from EFCRFNet “ECRF (No CA) “refers to the experimental results obtained by removing the CA module from ECRF.
Table 3.
‘No EFEM’ refers to the experimental results obtained after removing the EFEM module from EFCRFNet; ‘EFEM (No CA)’ refers to the experimental results obtained after removing the CA module from EFEM‘ EFEM (No SA) ‘refers to the experimental results obtained after removing the SA module from EFEM.
Fig 14.
Visual comparison of saliency maps generated by the baseline, EFEM only, ECRF only, and EFCRFNet.
Table 4.
Comparing FLOPs, Param, and FPS indicators between EFCRFNet and eight SOD methods.