mobilenetv3(width_mult, configs; max_width = 1024, nclasses = 1000)

Create a MobileNetv3 model. (reference).

Arguments

• width_mult: Controls the number of output feature maps in each block (with 1.0 being the default in the paper; this is usually a value between 0.1 and 1.4)
• configs: a “list of tuples” configuration for each layer that details:
  • k::Int - The size of the convolutional kernel
  • c::Float - The multiplier factor for deciding the number of feature maps in the hidden layer
  • t::Int - The number of output feature maps for a given block
  • r::Int - The reduction factor (>= 1 or nothing to skip) for squeeze and excite layers
  • s::Int - The stride of the convolutional kernel
  • a - The activation function used in the bottleneck (typically hardswish or relu)
• max_width: The maximum number of feature maps in any layer of the network
• nclasses: the number of output classes