All convolutions inside of a dense block are ReLU-activated and use batch normalization. Channel-sensible concatenation is simply possible if the height and width dimensions of the information keep on being unchanged, so convolutions inside of a dense block are all of stride 1. Pooling levels are inserted between dense https://financefeeds.com/bitget-wallet-announces-payfi-strategy/