在fullConnectedLayer之前如何将要素地图添加到CNN?

时间:2019-02-28 22:43:40

标签: machine-learning deep-learning

layers = [ imageInputLayer([width,height,channels]); %'DataAugmentation','none'); %'Normalization','none'); convolution2dLayer(2,16,'Padding',1) batchNormalizationLayer 替代 在这里添加功能图 fullyConnectedLayer(12) softmaxLayer 分类图层];

1 个答案:

答案 0 :(得分:0)

您的要素图(f1)尺寸应与卷积层输出要素图(cf2)匹配

Pytorch-torch.cat((f1,cf2),dim = 1) Tensorflow-tf.concat([t1,t2],axis = 1)

注意:输出将使过滤器数量增加一倍enter image description here

相关问题