A lightweight convolutional neural network as an alternative to DIC to measure in-plane displacement fields

Convolutional Neural Networks (CNNs) are now commonly used in the computer vision community, particular for optical flow estimation. Some attempts to use such tools measure displacement and strain fields from pairs of reference/deformed speckle images (like Digital Image Correlation) have been recently reported literature. The aim this work is twofold. first one customize a state-of-the-art CNN dedicated estimation reach better performance when processing images. This mainly obtained by removing deepest levels. second further simplify reducing as much possible number filters remaining levels while keeping equivalent metrological original version, order accelerate image on power-efficient compact Graphics Processing Unit (GPU). Synthetic deformed through suitable field assess different versions tested study. We focus sub-pixel part considered attempt, being more challenging determine than integer displacements at pixel scale. latter can be found cross-correlation or with rough version DIC. Real simplest results compared those provided classic subset-based Correlation. two main conclusions i- that customization procedure improves ii- ultimate simplified globally initial despite drastic simplification end procedure. lies between DIC first- second-order subset shape functions.