Unsupervised domain translation using invertible nets with strict MLE training. This project is currently under development.
The following methods have been implemented in my pytorch lightning setup:
Method | Link to project |
---|---|
CycleGAN | https://junyanz.github.io/CycleGAN/ |
CutGAN | http://taesung.me/ContrastiveUnpairedTranslation/ |
AlignFlow | https://github.com/ermongroup/alignflow |
Here are some results from a preliminary inverstigation on toy data. The input and outputs are 2D points and the mapping to be learnt is rather simple.