River Segmentation of Remote Sensing Images Based on Composite Attention Network

Table 3

The net performance comparison of different parameter weights.

(μ, λ)	PA	MPA	MIoU	FMIoU
(1.0, 1.0)	0.9703	0.9498	0.9432	0.9653
(1.2, 0.8)	0.9724	0.9537	0.9471	0.9682
(1.4, 0.6)	0.9735	0.9549	0.9489	0.9709
(1.6, 0.4)	0.9743	0.9556	0.9502	0.9728
(1.8, 0.2)	0.9728	0.9542	0.9476	0.9687

Bold indicates the maximum value of each evaluation index.