Change Detection of Remote Sensing Images Based on Attention Mechanism

Table 4

Comparison of F1, recall values, and FLOPS values with the stack autoencoder, U-Net, PSPNet, and ours on the test datasets.

Module	F1	Recall	Pixel accuracy	FLOPS × 10^9
Stack autoencoder	0.463	0.391	0.568	3.2
U-Net	0.689	0.691	0.687	5.6
PSPNet	0.724	0.704	0.745	11.2
Ours	0.743	0.697	0.796	4.5