The Project title is self-explanatory. We as humans possess the ability to draw important characteristics from a human action (or gesture) by looking at it once. We are also able to differentiate between several actions which have been shown to us only once. This ability is greatly attributed to our attention mechanism. We are able to focus on the moving objects and follow their trajectory to form an understanding of an action.
The project aims to develop an architecture that is able to perform One-Shot Classification of actions by incorporating attention into the architecture.