# ------------------------------------------------------------------------------# Copyright c Microsoft# Licensed under the MIT License# Create by Bin Xiao BinXiaomicrosoftcom# Modified by Tianheng Ch

The code defines a High-Resolution Network for facial landmark detection. The network consists of a stem net, followed by three stages of High-Resolution Modules. Each module contains multiple branches with different resolutions, which are fused together using a specified method. The output of the last stage is passed through a head consisting of a convolutional layer to produce the final landmark predictions. The code also includes functions for initializing the weights of the network and getting the network architecture