Abstract: A convolutional neural network model was proposed to predict the mechanical properties of Gd₂O₃/6061Al neutron shielding material. The EBSD microstructure of Gd₂O₃/6061Al neutron shielding material and its corresponding tensile properties were used as data set to train and verify the convolutional neural network model. The results show that using multiple microscopic images without any artificial image processing, the convolutional neural network can get good training results and its performance is better than that of traditional testing methods. The convolutional neural network can capture the existence of grains and some statistical informations of grains. There is a strong correlation between grain number and grain size.