Extraction and Classification of the Supervised Coastal Objects Based on HSRIs and a Novel Lightweight Fully Connected Spatial Dropout Network

Table 2

Results of the individual target’s effectiveness evaluation.

Metrics	Methods	Mangrove	Aquaculture raft	Aquaculture pond
PA (%)	U-Net	86.64	94.33	78.77
	U-Net (CBS)	94.88	96.80	89.77
	DeepLabV3+	94.50	94.95	90.71
	DeepLabV3+ (CBS)	94.04	94.40	88.90
	LFCSDN (ours, BN)	94.88	93.76	89.85
	LFCSDN (ours, CBS)	93.50	90.99	88.49
UA (%)	U-Net	88.98	89.10	56.26
	U-Net (CBS)	91.65	94.23	79.23
	DeepLabV3+	92.45	88.58	83.05
	DeepLabV3+ (CBS)	94.45	91.91	87.92
	LFCSDN (ours, BN)	91.06	94.32	82.19
	LFCSDN (ours, CBS)	92.74	94.25	83.63
F1 (%)	U-Net	87.80	91.64	65.64
	U-Net (CBS)	93.24	95.49	84.17
	DeepLabV3+	93.47	91.66	86.71
	DeepLabV3+ (CBS)	94.24	93.14	88.40
	LFCSDN (ours, BN)	92.93	94.04	85.85
	LFCSDN (ours, CBS)	93.12	92.59	85.99
IoU (%)	U-Net	78.25	84.58	48.86
	U-Net (CBS)	87.33	91.38	72.67
	DeepLabV3+	87.74	84.60	76.54
	DeepLabV3+ (CBS)	89.11	87.16	79.22
	LFCSDN (ours, BN)	86.79	88.74	75.20
	LFCSDN (ours, CBS)	87.13	86.21	75.43