Fig. 2From: Reverse active learning based atrous DenseNet for pathological image classificationExamples of atrous convolutions with different dilation rates. The purple squares represent the positions of kernel parametersBack to article page