In this study, several artificial neural network (ANN) models were developed with the help of microstructural characteristics to predict uniaxial compressive strength (UCS) and elastic modulus (E) of migmatites. To this end, 51 migmatite samples were prepared and for each sample some microstructural characteristics along with UCS and E were determined. A semi-automatic technique was implemented to quantify twenty microstructural characteristics including mineral size (area, perimeter, equivalent circular diameter, minimum Feret diameter, maximum Feret diameter), mineral shape (aspect ratio, orientation, compactness, roundness, rectangularity, solidity, convexity, concavity, form factor), fabric coefficients (index of interlocking, index of grain size homogeneity), mineral contents (quartz content, feldspar content), and mineralogical indices (saturation index, and coloration index) based on digital imaging of representative parts of thin sections of samples. Area, orientation, concavity, index of interlocking, feldspar content and saturation index were chosen as inputs of the prediction models. Here, 6, 10 and 10 ANN models were developed with 5, 4 and 3 inputs, respectively, using these microstructural characteristics. Bayesian regularization was employed for training the models to improve the generalization capacity of the models. From these models, the model with orientation, index of interlocking and feldspar content as inputs was selected for producing prediction charts. For this model, mean squared error and correlation coefficient were obtained as 0.0148 and 0.854 for training data, and 0.0143 and 0.860 for test data, respectively. The proposed models and prediction charts in this study can show high prediction capability, given that they are used for similar types of rocks.