All convolutions in the dense block are ReLU-activated and use batch normalization. Channel-wise concatenation is barely attainable if the height and width Proportions of the information continue to be unchanged, so convolutions inside of a dense block are all of stride one. Pooling layers are inserted between dense blocks https://financefeeds.com/four-copyright-firms-secure-mica-licenses-in-the-netherlands/