All convolutions in a very dense block are ReLU-activated and use batch normalization. Channel-intelligent concatenation is simply attainable if the height and width dimensions of the information keep on being unchanged, so convolutions within a dense block are all of stride 1. Pooling layers are inserted between dense blocks https://financefeeds.com/ig-group-reports-11-revenue-growth-and-41-profit-rise-in-h1-results/