All convolutions within a dense block are ReLU-activated and use batch normalization. Channel-wise concatenation is only probable if the peak and width dimensions of the data keep on being unchanged, so convolutions inside a dense block are all of stride one. Pooling layers are inserted between dense blocks for https://financefeeds.com/cls-global-admits-to-fraud-in-u-s-undercover-copyright-investigation/