Fig. 4From: Deep semantic learning for acoustic scene classificationCompare the performance of kernel size \(2 \times 3\) and \(3 \times 3\) in channel-wise 80 to 100. Left: the feature map with \(3 \times 3\) kennel size of the Mini-SegNet encoder output. Right:the feature map with \(2 \times 3\) kennel size of the Mini-SegNet encoder outputBack to article page