DiffeomorphicPIV

This repository contains code for the accepted paper Diffeomorphic Particle Image velocimetry. In this work, a diffeomorphic PIV technique is proposed to reduce the curvature effect of the non-straight particle trajectory. Different from other existing PIV techniques, our diffeomorphic PIV computes the real curved particle trajectory to achieve accurate velocity measurement.

Motivation

The diffeomorphic PIV uses the curved trajectory (streamline of velocity field) to explain the image displacement between TWO recordings. It is significantly different from the straight-line approximation of existing PIV techniques. More info is referred to the paper.

Install dependencies

conda install pytorch==1.7.0 torchvision==0.8.0 torchaudio==0.7.0 cudatoolkit=11.0 -c pytorch
conda install numpy matplotlib opencv seaborn
conda install -c conda-forge openpiv
conda install -c conda-forge cupy

Download the well-trained deep model for DeepPIV

cd UnFlowNet
wget https://github.com/erizmr/UnLiteFlowNet-PIV/raw/master/models/UnsupervisedLiteFlowNet_pretrained.pt

The experiments

• Exp1.ipynb: Investigate the converge performance w.r.t the iteration number
• Exp2.ipynb: Test on 3 Lamb-Oseen flows and 3 Sin flows;
• Exp3.ipynb: Test on 3 real PIV cases;

BibTeX

@article{lee2021diffeomorphic,
  author={Lee, Yong and Mei, Shuang},  
  journal={IEEE Transactions on Instrumentation and Measurement},  
  title={Diffeomorphic Particle Image Velocimetry},  
  year={2021},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TIM.2021.3132999}}

Questions?

For any questions regarding this work, please email me at yongli.cv@gmail.com.

Acknowledgements

Parts of the code/deep net in this repository have been adapted from the following repos:

• OpenPIV/openpiv-python
• erizmr/UnLiteFlowNet-PIV
• opencv/opencv