PIV-FlowDiffuser:Transfer-learning-based denoising diffusion models for PIV

TL;DR

This work addresses the domain gap between synthetic training data and real PIV imagery by introducing PIV-FlowDiffuser, a transfer-learning-enabled denoising diffusion model. It adapts FlowDiffuser to PIV through a dual-encoder architecture and a conditional recurrent denoising decoder, with an upsampling strategy to capture small-scale turbulence, and pre-trains on large optical-flow datasets before fine-tuning on synthetic PIV data. The approach yields a 59.4% reduction in average end-point error over a strong RAFT-based baseline on Cai's dataset and demonstrates superior generalization to out-of-domain and practical PIV images, while incurring modest computational costs thanks to transfer learning. Overall, the method highlights the potential of combining denoising diffusion models with transfer learning to achieve accurate, robust PIV measurements under challenging turbulent conditions.

Abstract

Deep learning algorithms have significantly reduced the computational time and improved the spatial resolution of particle image velocimetry~(PIV). However, the models trained on synthetic datasets might have a degraded performance on practical particle images due to domain gaps. As a result, special residual patterns are often observed for the vector fields of deep learning-based estimators. To reduce the special noise step-by-step, we employ a denoising diffusion model~(FlowDiffuser) for PIV analysis. And the data-hungry iterative denoising diffusion model is trained via a transfer learning strategy, resulting in our PIV-FlowDiffuser method. Specifically, (1) pre-training a FlowDiffuser model with multiple optical flow datasets of the computer vision community, such as Sintel, KITTI, etc; (2) fine-tuning the pre-trained model on synthetic PIV datasets. Note that the PIV images are upsampled by a factor of two to resolve the small-scale turbulent flow structures. The visualized results indicate that our PIV-FlowDiffuser effectively suppresses the noise patterns. Therefore, the denoising diffusion model reduces the average end-point error~($AEE$) by 59.4% over RAFT256-PIV baseline on the classic Cai's dataset. Besides, PIV-FlowDiffuser exhibits enhanced generalization performance on unseen particle images due to transfer learning. Overall, this study highlights the transfer-learning-based denoising diffusion models for PIV. And a detailed implementation is recommended for interested readers in the repository https://github.com/Zhu-Qianyu/PIV-FlowDiffuser.

Figures (5)

• Figure 1: Results of the RAFT256-PIV method lagemann2021deep on two test cases. The left gives the vector fields, while the right part presents corresponding error maps. Note that some special error patterns are observed in the residuals, which could be further reduced with noise removal.
• Figure 2: (a) The FlowDiffuser luo2024flowdiffuser includes two encoders (basic encoder and context encoder) and a series of conditional recurrent denoising decoder (RDD). (b) The training method for PIV-FlowDiffuser. The initial weights are from the pre-trained model, and a simple adaptation module (scale up & scale down) is adopted to better predict small-scale turbulence. The entire model was subsequently fine-tuned using a PIV-specified dataset.
• Figure 3: Velocity fields and corresponding absolute residuals of Problem Class 1 computed by different methods. Two cases (left: JHTDB, right: SQG) are considered. The color backgrounds denote the corresponding velocity/residual magnitude. Best viewed in color. (unit: pixels per frame)
• Figure 4: Velocity fields and corresponding absolute residuals computed by different methods. Two cases (left: a JHTDB from Problem Class 1, right: a JHTDB from Problem Class 2) are considered. The color backgrounds denote the corresponding velocity/residual magnitude. Best viewed in color. (unit: pixels per frame)
• Figure 5: Two cases of TWCF data are visualized with separate velocity components. The color backgrounds denote the corresponding component value. Best viewed in color. (unit: pixels per frame)