Pyramidal deep-learning network for dense velocity field reconstruction in particle image velocimetry

Zhang, Wei; Nie, Xiangyu; Deng, Xue; Sun, Zhiwei

doi:10.1007/s00348-022-03540-4

Cited by 8 publications

(1 citation statement)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These methods formulate their cost functions by incorporating photometric loss, flow smoothness loss, and consistency loss. Furthermore, Zhang et al [19] introduced UnPWCNet-PIV, which builds upon PIV-PWCNet [20]. Other notable deep learning-based algorithms for flow estimation in PIV include CC-FCN by Gao et al [21], which synergistically combines cross-correlation and fully convolutional network approaches; CascLiteFlowNet-R-en by Guo et al [22], a novel cascaded CNN tailored for time-resolved PIV (TR-PIV) inspired by LiteFlowNet; Yu et al developed Deep-TRPIV [23], a multi-frame architecture for optical flow prediction from successive particle images, drawing inspiration from RAFT architecture; ARaft-FlowNet by Han and Wang [24] and DeepST-CC by Yu et al [25] leverage the RAFT optical flow model, employing attention-based architectures to improve tracer particle motion recognition.…”

Section: Introductionmentioning

confidence: 99%

Experimental dataset investigation of deep recurrent optical flow learning for particle image velocimetry: flow past a circular cylinder

Anjaneya Reddy,

Wahl,

Sjödahl

2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Current optical flow-based neural networks for Particle Image Velocimetry (PIV) are largely trained on synthetic datasets emulating real-world scenarios. While synthetic datasets provide greater control and variation than what can be achieved using experimental datasets for supervised learning, it requires a deeper understanding of how or what factors dictate the learning behaviors of deep neural networks for PIV. In this study, we investigate the performance of the Recurrent All-Pairs Field Transforms (RAFT-PIV) network, the current state-of-the-art deep learning architecture for PIV, by testing it on unseen experimentally generated datasets. The results from RAFT-PIV are compared with a conventional cross-correlation-based method, Adaptive PIV. The experimental PIV datasets were generated for a typical scenario of flow past a circular cylinder in a rectangular channel. These test datasets encompassed variations in particle diameters, particle seeding densities, and flow speeds, all falling within the parameter range used for training RAFT-PIV. We also explore how different image pre-processing techniques can impact and potentially enhance the performance of RAFT-PIV on real-world datasets. Thorough testing with real-world experimental PIV datasets reveals the resilience of the optical flow-based method’s variations to PIV hyperparameters, in contrast to the conventional PIV technique. The ensemble-averaged Root Mean Squared Errors (RMSE) between the RAFT-PIV and Adaptive PIV estimations generally range between 0.5 to 2 [px] and show a slight reduction as particle densities increase or Reynolds numbers decrease. Furthermore, findings indicate that employing image pre-processing techniques to enhance input particle image quality doesn't improve RAFT-PIV predictions; instead, it incurs higher computational costs and impacts estimations of small-scale structures.

show abstract