Dynamic Multiscale Feature Fusion Method for Underwater Target Recognition

Table 2

Conventional underwater image target recognition accuracy and time.

Method	Torpedo	Torpedowake	Submarine	Frogman	Bubble	AUV	mAP	Time
FISHnet	0.5711	0.6477	0.9286	0.7553	0.6944	0.9197	0.7528	0.398
SiamFPN	0.4376	0.4163	0.9294	0.7302	0.6419	0.9283	0.6806	0.231
SA-FPN	0.4742	0.6372	0.8284	0.7307	0.7017	0.9300	0.7170	0.245
Literature [33]	0.6011	0.5627	0.7855	0.7614	0.7765	0.9056	0.7321	0.100
Ours	0.6715	0.5449	0.9573	0.7468	0.7645	0.8732	0.7597	0.223

Note: the bold value in the table indicates excellent indicators of each algorithm.