I need some help with understanding the backward warping in paper Neural Supersampling for Real-time Rendering.
In section 3.1 They state "First, it (referring t
