Abstract: In order to generate controllable porous structure inside different models,we present a method of transferring the two-dimensional porous pattern to three-dimensional structures in this paper.The porous pattern is consisted of the distribution pattern and filling pattern.A convolutional neural network is used to memorize the distribution pattern,and a pattern transferring process is applied to transfer the two-dimensional distribution to three-dimensional models.Experiments verified that our method can provide an effective mechanism for porous structure design.