All convolutions in a dense block are ReLU-activated and use batch normalization. Channel-sensible concatenation is just doable if the peak and width dimensions of the info stay unchanged, so convolutions inside a dense block are all of stride 1. Pooling levels are inserted between dense blocks for even further https://financefeeds.com/why-shiba-inu-holders-are-rushing-to-buy-into-the-new-viral-altcoin-remittix-as-ico-raises-1m-in-under-24-hours/