A framework for multi-task learning, where you may precondition tasks and compose them into bigger tasks. Conditional objectives and per-task evaluations and interpretations.
Why do you think that https://github.com/avinashpaliwal/Super-SloMo is a good alternative to dnn.cool