spatial

Definition

Concerning position and layout — the height and width axes of an image, or where things sit in a 3D scene. It comes up when discussing inductive bias: convolutions bake in locality and translation equivariance, while a vision transformer must learn spatial structure from position embeddings over patches. For video the term pairs with temporal, giving spatio-temporal modelling.

Explainers using this term

• Paper Explained: SpatialBlock — Teaching Vision-Language Models Spatial Sense with 15,000 Block-Stacking PuzzlesSpatialBlock: Enhancing Spatial Intelligence in LVLMs via Synthetic Block-Stacking Problem
• Paper Walkthrough: CosmoH2G — Transferring a Human Hand's Flips and Rotations to a Two-Finger GripperCosmoH2G: A Hand-to-Gripper Transfer Dataset and Baseline Method for Object Manipulation with Complex Spatial Movements
• Paper Walkthrough — UrbanGround: Where MLLM Agents Break Down on a Real StreetUrbanGround: From Local Perception to Spatial Agency in a Real-Scale City