Fig. 2.

A diagram illustrating the architecture of basic building blocks and the architecture of different models. (a) The architecture of ResNet building block. (b) The architecture of SE-ResNet building block. (c) The architecture of SE-ResNeXt building blocks. We used eight transformation layers along with 32 filters for each transformation layer. (d) The architecture of different models. Despite the difference in building blocks, the architectures of the ResNet, SE-ResNet, and SE-ResNeXt models were the same. BN = batch normalization, GAP = global average pooling, Conv = convolutional, Avg = average.