Dual-Path Residual “Shrinkage” Network for Side-Scan Sonar Image Classification

Table 9

Number of correct and incorrect predictions of drowning victim category when the number of samples is 5, 10, 15, and 18.

Number of samples	Predicted class	Model
		ResNet34 + SE	ResNet34 + CBAM	ResNet34 + RSBU-CW	Proposed
5	Drowning victim	5	5	5	5
	Aircraft	0	0	0	0
	Seafloor	0	0	0	0
	Shipwreck	0	0	0	0
10	Drowning victim	9	9	9	9
	Aircraft	0	0	0	0
	Seafloor	0	0	0	0
	Shipwreck	1	1	1	1
15	Drowning victim	12	11	12	14
	Aircraft	0	1	1	0
	Seafloor	0	0	0	0
	Shipwreck	3	3	2	1
18	Drowning victim	13	10	14	14
	Aircraft	0	6	0	1
	Seafloor	1	0	0	0
	Shipwreck	4	2	4	3