The branch of mathematics concerned with the properties of space and of figures within it — points, lines, curves, surfaces, and solids — and with the notions of distance, angle, area, volume, and transformation that relate them. From Euclid’s axiomatic plane geometry through analytic, projective, differential, and computational branches, geometry supplies the formal language in which spatial computing describes the world: coordinate frames and transformations in robotics, projective camera models in computer vision, meshes and parametric surfaces in graphics, and spatial subdivisions such as Voronoi diagrams in computational geometry.
Semantic Classification
Content
Definition
Geometry is among the oldest of mathematical disciplines and the one most directly concerned with space. Euclid’s Elements (c. 300 BCE) organised it axiomatically: from a handful of postulates about points, lines, and circles, theorems about congruence, similarity, and measurement follow by deduction. Descartes’ analytic geometry fused it with algebra by assigning coordinates to points, so that curves become equations and geometric questions become computations — the move that ultimately made geometry programmable. Subsequent centuries diversified the field: projective geometry studies properties invariant under perspective projection, differential geometry treats smooth curves and surfaces with calculus, and non-Euclidean geometries showed that the parallel postulate is a choice, not a necessity — a result that later underpinned general relativity.
Felix Klein’s Erlangen programme (1872) gave the modern unifying view: a geometry is characterised by its group of transformations and the properties that group leaves invariant. Euclidean geometry preserves distances; affine geometry preserves parallelism; Projective Geometry preserves incidence and cross-ratio. This transformation-centred perspective is precisely the one spatial computing inherits — rigid-body transforms chain robot joint frames, homogeneous coordinates carry projective camera models, and quaternions parameterise rotation — with Linear Algebra serving as the computational engine throughout.
In this knowledge graph, geometry is the foundation beneath its applied descendants. Computational Geometry turns geometric questions into algorithms and complexity bounds — convex hulls, triangulations, and spatial structures such as the Voronoi Diagram. Scene Geometry denotes the recovered geometric structure of real environments in vision and neural rendering. Neighbouring Topology studies the properties that survive continuous deformation, where geometry’s metric notions are deliberately forgotten.
Technical Details
-
Primitives and measures: points, lines, planes, conics, polygons, polyhedra, and smooth manifolds; distance, angle, area, volume, curvature.
-
Representations in computing: homogeneous coordinates unify rotation, translation, and projection as matrix products; boundary representations, polygon meshes, implicit surfaces (signed distance fields), and parametric splines (NURBS) encode shape for graphics, CAD, and simulation.
-
Transformation hierarchy: Euclidean ⊂ similarity ⊂ affine ⊂ projective — each step preserves fewer invariants, and computer vision routinely works at the projective level before upgrading reconstructions to metric.
-
Algorithmic layer: geometric predicates (orientation, in-circle) demand careful numerical treatment; robustness under floating-point arithmetic is a defining concern of computational geometry libraries such as CGAL.
Current Landscape
-
CGAL, the reference computational-geometry library, published version 6.1 in 2025 — adding constrained 3D Delaunay triangulations, self-intersection repair via 3D snap rounding, and a 10x acceleration of mesh clipping/splitting — followed by version 6.2 on 11 June 2026, which introduced 2D alpha wrapping, 3D generalised barycentric coordinates, and a restructured Polygon Mesh Processing suite.
-
Google DeepMind’s AlphaGeometry2 (February 2025) raised automated solving of IMO geometry problems from 54% to 84% over the 2000–2024 set — surpassing the average IMO gold medallist — by pairing a Gemini-based language model with a faster symbolic engine and a shared-knowledge ensemble tree search; it formed part of the system that achieved silver-medal standard at IMO 2024.
-
Geometric deep learning continues to industrialise geometry within machine learning: libraries such as PyTorch Geometric treat learning on meshes, point clouds, and molecular structures as geometry-aware computation over irregular domains.
Sources: