space-time

Definition

Jointly over space and time — the defining structure of video, where two spatial axes are joined by a temporal one. Modeling it means capturing both appearance within a frame and motion across frames, which is what space-time attention does. Attending over all positions and all frames at once costs quadratic compute in the total number of tokens, so most practical architectures factorize the operation into separate spatial and temporal passes.

Explainers using this term

• Video Understanding from Scratch — From a Pile of Frames to a Sense of TimeIs Space-Time Attention All You Need for Video Understanding? (TimeSformer)