A compact HRNet implementation in PyTorch without YAML configuration
from hrnet import HRNet
'''
parameters from left to right:
input channels, first branch channels (hyper parameter), output channels
notice:
Only the number of first branch channels is necessary, numbers of channels of
other branches are calculated according to the paper
'''
model = HRNet(3, 16, 8)torch==1.5.0