Holy crap this is awesome

I think the getting started bit (manual, videos etc.) jumps too far into the details of the implementation. It assumes the users are familiar with all the concepts already, even though it is the "getting started" section. It might pick up more users by starting with a basic challenge and explain how you would attack it using deep learning in general. By first explaining some of the basic concepts of deep learning and showing how those can be applied to solve a task, and *then* show how we can realize this on a given set of data with your toolkit it is much easier for new (to deep learning as well) users to understand how we can use it to attack concrete challenges. The more advanced example can then build on that, but should have similar examples that contain the data to train the models and test the result, or describe what data needs to be generated to do that.

Download doesn't work - project dead?

Looks great

Looks great!

Promo-SM

Promo_SM 😚