Change Detection of Remote Sensing Images Based on Attention Mechanism

Table 1

The architecture details for our encoder.

Layer	Output size	Encoder
Conv1	512 × 512	7 × 7, 64, stride 2
Max pooling	256 × 256	3 × 3 stride 2
Residual unit	256 × 256
Attention module	256 × 256	Attention × 1
Residual unit	128 × 128
Attention module	128 × 128	Attention × 1
Residual unit	64 × 64
Attention module	64 × 64	Attention × 1
Residual unit	32 × 32