Predicting Micromechanical Properties of Cement Paste from Backscattered Electron (BSE) Images by Computer Vision

•A Residual Convolutional Neural Network (Res-Net) model was employed to predict indentation modulus and hardness of cement paste from BSE images.•A dataset comprising 40,000 nanoindentation results and 40,000 BSE images was built to train the model.•The resulting Res-Net model can precisely predict the elastic modulus and hardness with a R2 of 0.85 and 0.88.•Deconvolution of Res-Net prediction obtains similar invariant statistics as tested by experiments, which gives strong evidence of the applicability of the Res-Net model.