PyConvU-Net

A lightweight and multiscale network for biomedical image segmentation

Introduction

dataset.py is used to load dataset and preprocess the data

flops_counter.py is to calculate the number of parameters and computational complexity

metrics.py is to calculate the metrics: MIoU and Dice

plot.py is to plot the loss, MIoU and Dice curve

Environment

Pytorch

Run

python main.py --arch xxx --dataset xxx --epoch xxx --batch_size xxx

Reference

https://github.com/Andy-zhujunwen/UNET-ZOO

https://github.com/iduta/pyconv

Citations

@article{li2021pyconvu,
  title={PyConvU-Net: a lightweight and multiscale network for biomedical image segmentation},
  author={Li, Changyong and Fan, Yongxian and Cai, Xiaodong},
  journal={BMC bioinformatics},
  volume={22},
  number={1},
  pages={1--11},
  year={2021},
  publisher={Springer}
}