Structure of EfficientNet-B1. Conv in the figure denotes the convolution operation and s denotes the step size of the convolution. For example, ‘Conv3*3, s1’ denotes a convolution operation with a convolution kernel size of 3*3 and a step size of 1. BN denotes batch normalization, which serves to batch normalize the current input and normalize the data under the same scale to speed up model training. Swish refers to the swish activation function, whose expression is y  = x * sigmoid(x).