Actual numeric convolution example with strides and padding using convolution2dlayer command

For convolution2dlayer command , how do I simulate an actual numeric example of convolution with strides and padding ?

Answer 1

[Dinesh Yadav]

Hi, go through the following code

input = randi(10,10,10);      %creating a random input 2-d array  
layer=imageInputLayer([10 10],'Name','Input','Normalization','none');   % defining input layer of network
layer2 = convolution2dLayer(3,1,'Padding',[1 1 1 1],'Stride',2,'WeightsInitializer',...   
    @(sz) rand(sz) * 0.0001,'BiasInitializer', @(sz) rand(sz) * 0.0001,'Name','con');  % defining convolution layer of network
layer1=[layer layer2];
lg=layerGraph(layer1);
net=dlnetwork(lg);
out=predict(net,dlarray(input,'SSC'))

As input matrix in 10x10 and convolution is with padding 1 and stride 2 output matrix is 5x5.

Hope it helps.