CBAM_MatLab/Channel Attention Module at main · ImageVisioner/CBAM_MatLab · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
lgraph = layerGraph();
tempLayers = imageInputLayer([227 227 3],"Name","imageinput");
lgraph = addLayers(lgraph,tempLayers);

tempLayers = maxPooling2dLayer([5 5],"Name","maxpool_1","Padding","same");
lgraph = addLayers(lgraph,tempLayers);

tempLayers = averagePooling2dLayer([5 5],"Name","avgpool2d_1","Padding","same");
lgraph = addLayers(lgraph,tempLayers);

tempLayers = [
    multiplicationLayer(2,"Name","multiplication_1")
    fullyConnectedLayer(10,"Name","fc_1")
    fullyConnectedLayer(10,"Name","fc_2")
    fullyConnectedLayer(10,"Name","fc_3")
    leakyReluLayer(0.01,"Name","leakyrelu_1")];
lgraph = addLayers(lgraph,tempLayers);

tempLayers = averagePooling2dLayer([5 5],"Name","avgpool2d_2","Padding","same");
lgraph = addLayers(lgraph,tempLayers);

tempLayers = maxPooling2dLayer([5 5],"Name","maxpool_2","Padding","same");
lgraph = addLayers(lgraph,tempLayers);

tempLayers = [
    multiplicationLayer(2,"Name","multiplication_2")
    leakyReluLayer(0.01,"Name","leakyrelu_2")
    sigmoidLayer("Name","sigmoid")
    yolov2OutputLayer([16 16;32 32],"Name","yolov2-out")];
lgraph = addLayers(lgraph,tempLayers);

% 清理辅助变量
clear tempLayers;

%链接所有层
lgraph = connectLayers(lgraph,"imageinput","maxpool_1");
lgraph = connectLayers(lgraph,"imageinput","avgpool2d_1");
lgraph = connectLayers(lgraph,"maxpool_1","multiplication_1/in2");
lgraph = connectLayers(lgraph,"avgpool2d_1","multiplication_1/in1");
lgraph = connectLayers(lgraph,"leakyrelu_1","avgpool2d_2");
lgraph = connectLayers(lgraph,"leakyrelu_1","maxpool_2");
lgraph = connectLayers(lgraph,"avgpool2d_2","multiplication_2/in1");
lgraph = connectLayers(lgraph,"maxpool_2","multiplication_2/in2");