Search for a command to run...
Knowing What, Where and When to Look: Efficient Video Action Modeling with Attention