基于混合注意力机制和Deeplabv3+的遥感影像建筑物提取方法

刘晨晨, 葛小三, 武永斌, 余海坤, 张蓓蓓

A method for information extraction of buildings from remote sensing images based on hybrid attention mechanism and Deeplabv3+

LIU Chenchen, GE Xiaosan, WU Yongbin, YU Haikun, ZHANG Beibei

表1 CAM-Resnet50网络结构

Tab.1 CAM-Resnet50 network structure

层级名称	输入尺寸	操作模块	重复次数	输出通道数
Conv1	224²×3	7×7卷积	1	128
Conv2_x	112²×64	Conv Block	1	256
Conv2_x	112²×64	Identity Block	2	256
Conv3_x	56²×256	Conv Block	1	512
Conv3_x	56²×256	Identity Block	3	512
Conv4_x	28²×512	Conv Block	1	1 024
Conv4_x	28²×512	Identity Block	5	1 024
Conv5_x	14²×1 024	Conv Block	1	2 048
Conv5_x	14²×1 024	Identity Block	2	2 048
Conv6	7²×2 048	1×1卷积	1	2 048