Deep De-Homogenization: pretrained models

Data related to the pretrained models used in: Elingaard, M. O., Aage, N., Bærentzen, J. A., & Sigmund, O. (2022). De-homogenization using convolutional neural networks. Computer Methods in Applied Mechanics and Engineering, 388, 114197. https://doi.org/10.1016/j.cma.2021.114197The directory contains two different models, one for a frequency of 10 pixels/period, and one for a frequency of 20 pixels/period. These have been denoted step2. For completion the weights used to initialize the training for the second step of the algorithm have also been included and are denoted step1. Files are saved in the .pth format, as recommended by PyTorch, and can be loaded using torch.load() or model.load​_state_dict(), see https://pytorch.org/tutorials/beginner/saving_loading_models.html for more information.