A lightweight neural network designed for fluid velocimetry

Abstract

We devise a novel lightweight image matching architecture (), which is designed and optimized for particle image velocimetry (PIV). is a convolutional neural network (CNN) that performs symmetric image matching and employs an iterative residual refinement strategy, which allows us to optimize the total number of refinement steps to balance accuracy and computational efficiency. The network is trained on kinematic datasets with a loss function that penalizes larger gradients. We consider a six-level () and a four-level () version of the network and demonstrate that they are considerably leaner and faster than a state-of-the-art network designed for optical flow. reconstructs the velocity field from synthetic and experimental PIV images with an accuracy comparable or superior both to existing CNNs as well as to state-of-the-art cross-correlation methods (i.e., a commercial implementation of ). Although less accurate, allows a significant reduction of the computational costs with respect to any other method considered. All CNNs prove more robust than with respect to particle loss and allow effective error reduction by increasing the particle seeding density. Thanks to reduced computational cost and memory requirement, we envision the deployment of on low-cost devices to provide affordable, real-time inference of the flow field during PIV measurements.