README.md

HRNet

A compact HRNet implementation in PyTorch without YAML configuration

Performance

Usage

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)

Requirements

torch==1.5.0

References

• Deep High-Resolution Representation Learning for Visual Recognition, 2020
• Official implementation