How can I make Grad-cam of CNN net

조회 수: 2 (최근 30일)
Jaewon Kim
Jaewon Kim 2019년 8월 18일
답변: Kenta 2020년 7월 11일
I made CNN net based on 1-D data (1*204)
layers=[
imageInputLayer([1 204]);
convolution2dLayer([1 3],64,'Stride',1,'Padding',[0 1]);
convolution2dLayer([1 3],64,'Stride',1,'Padding',[0 1]);
batchNormalizationLayer
reluLayer();
maxPooling2dLayer([1 2],'Stride',[1 2]);
convolution2dLayer([1 3],128,'Stride',1,'Padding',[0 1]);
convolution2dLayer([1 3],128,'Stride',1,'Padding',[0 1]);
batchNormalizationLayer
reluLayer();
maxPooling2dLayer([1 2],'Stride',[1 2]);
convolution2dLayer([1 3],256,'Stride',1,'Padding',[0 1]);
convolution2dLayer([1 3],256,'Stride',1,'Padding',[0 1]);
batchNormalizationLayer
reluLayer();
maxPooling2dLayer([1 2],'Stride',[1 2]);
fullyConnectedLayer(400);
fullyConnectedLayer(400);
fullyConnectedLayer(4);
softmaxLayer();
classificationLayer();
];
I get a good CNN net, But I would like to which part of data is important to decide result.
I think It will be good way to use Grad-cam.
Many examples of Grad-cam are focused on Image. But my case is 1-D data not Image file..
Can you tell me which method will be good in my case?
Can I just use Grad-cam?,or more easy way to check which part is critical on result?

채택된 답변

Kenta
Kenta 2020년 7월 11일

추가 답변 (0개)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by