For Vision Datasets, it would be good to add stacked-mnist which is an mnist dataset with three channels. It is very useful in the field of GANs for analyzing mode collapse issues, as this dataset has 1000 modes and it is easy to evaluate whether there is a collapse or not.
For Vision Datasets, it would be good to add stacked-mnist which is an mnist dataset with three channels. It is very useful in the field of GANs for analyzing mode collapse issues, as this dataset has 1000 modes and it is easy to evaluate whether there is a collapse or not.