Microvec Introduces New AI PIV 2.0 Software

Microvec has introduced a new and improved AI Particle Image Velocimetry software that uses new techniques to get past the limitations in the previous version. End Quote

(1888PressRelease) June 24, 2026 - Microvec, one of the leaders in the field of fluid mechanics, is introducing an updated Artificial Intelligence Particle Image Velocimetry (AI PIV) 2.0 software at the XXVII Fluid Mechanics Conference, which is taking place from June 23 to 26, 2026 in Łódź, Poland. This is the second version of the AI PIV software originally introduced by Microvec in 2019. Microvec is once again introducing a unique and innovative product in the field of Particle Image Velocimetry.

The software is based on a deep learning technique called Recursive Prediction-Refinement Neural Network (RPR-NN), which Microvec has applied to fluid mechanics and particle image velocimetry. As is the way with AI, the results are only as good as the training data set, and the first software version introduced seven years ago was limited by the lack of diverse data sets. This new version has been improved by adding to and enhancing the data sets and training methods. In addition, an optical flow neural network commonly used in computer vision systems was added. Microvec’s AI PIV is based on an optical flow neural network where the global and quantitative velocity field can be extracted from images with improved computational efficiency without a reduction in accuracy.

This software is still the world’s only implementation of AI in PIV. The experimental results produced by the newer software version indicate that, compared to the previous AI PIV model, the incorporation of the proposed RPR-NN technique shows large improvements in the accuracy of the results.

“Microvec’s new AI PIV software has several advantages over the previous AI PIV version,” said Dr. Runjie Wei, CTO of Microvec. “It overcomes the issue of training on ideal images and the resulting discrepancy with images obtained under realistic experimental conditions with complex noise characteristics and non-ideal particle distribution. With the use of conventional optical flow solvers and spatial attention mechanisms, the final results show a much more accurate dense velocity field.”

###

space

Other Related Press Releases