PyTorch implementation of the example for the paper:

Alex Kendall, Yarin Gal, and Roberto Cipolla. "Multi-task learning using uncertainty to weigh losses for scene geometry and semantics." In CVPR 2018. [arXiv].

Some modifications were made based on the reference code, to make it more consistent with the original formula: https://github.com/Hui-Li/multi-task-learning-example-PyTorch