A Transformer-Based Network for Change Detection in Remote Sensing Using Multiscale Difference-Enhancement

Table 1

The average quantitative results of different CD methods on two datasets.

Method		STANet	SNUNet	BIT	ChangeFormer	TUNetCD (ours)
LEVIR-CD	Precision (%)	83.81	89.18	89.24	92.05	93.23
	Recall (%)	91.00	87.17	89.37	88.80	91.98
	F1 (%)	87.26	88.16	89.31	90.40	92.36
	IoU (%)	77.40	78.83	80.68	82.48	86.22
DSIFN-CD	Precision (%)	67.71	60.60	68.36	88.48	88.94
	Recall (%)	61.68	72.89	70.18	84.94	85.12
	F1 (%)	64.56	66.18	69.26	86.67	87.95
	IoU (%)	47.66	49.45	52.97	76.48	76.98