SciGraphs：Blender 导入导出插件

SciGraphs

SciGraphs is an open-source extension that embeds scientific graph processing directly within the modern 3D graphics environment of Blender. It turns abstract networks and real-world spatial graphs into native Blender objects, mapping analytical results (centralities, communities, topological invariants) to named attributes that drive color, scale, and animation — combining the analytical power of NetworkX, igraph and Graphviz with two ways of drawing the result: Geometry Nodes for Cycles and EEVEE, or the project's own GPU render engine.

SciGraphs supports two complementary domains:

• Combinatorial graphs, where classical network metrics, community structure, and algorithmic layouts are computed and mapped procedurally via attribute-driven properties.
• Spatial and geometric graphs, where networks are visualized using their intrinsic coordinate data (geospatial, sparse-matrix, mesh) or laid out via topological analysis, including multi-layer urban morphologies through the City2Graph module.

By unifying network analysis and visualization within Blender, SciGraphs enables reproducible 3D visualizations that extend network-geometry communication beyond the constraints of traditional 2D canvases.

Components included

• Data Import: GEXF, SuiteSparse Matrix Collection (.mtx), CSV edge lists and node-attribute tables, and live SQL databases (PostgreSQL, MySQL/MariaDB, SQLite; SQL Server optional). A time column can aggregate the import or restrict it to a range.
• OSMnx Street Networks: download by place, address, bounding box, point + radius, polygon or local .osm XML; routing and shortest paths, accessibility, edge speeds & travel times, centrality, bearings and orientation (rose diagrams), plus elevation/terrain (DEM) and basemap texturing.
• City2Graph (Urban Morphology & Transport): Overture Maps import (buildings, places, segments, connectors, land, water), morphological graphs (tessellation + private/public relations), proximity graphs, and GTFS transit analysis (DuckDB-backed) with travel-summary and origin–destination graphs and metapaths.
• Layouts: force-directed (NetworkX spring, igraph Fruchterman–Reingold, Kamada–Kawai, DrL, LGL, Davidson–Harel, Graphopt, ForceAtlas2, Yifan Hu), spectral (graph-Laplacian eigenvectors), classical MDS, Circle Packing for planar graphs (Collins–Stephenson), geometric 3D distributions (Fibonacci sphere, helix, spiral, cube), and directed/hierarchical layouts (Sugiyama layered, circular hierarchy). Graphviz engines (dot, neato, fdp, sfdp, twopi, circo, osage, patchwork) run via the bundled scigraphs-utils bindings. Every algorithm reports the backend it used and any fallback it took, so Circle Packing says whether the circles came out tangent. A Network Splitter decomposes layouts into Z-layers by community, degree, centrality, component or a custom expression. A layout can also be simulated frame by frame, its positions held in a GPU texture so playback never rebuilds the mesh.
• Analysis: centrality for undirected (degree, betweenness, closeness, eigenvector) and directed graphs (PageRank, HITS hub/authority, in/out-degree, Katz), computed through igraph or rustworkx where they are available and NetworkX otherwise; community detection with seven SurpriseMe algorithms (CPM, Infomap, RB, RN, RNSC, SCluster, UVCluster) and the Surprise quality metric via the bundled pysurprise bindings; directed-structure detection (DAG/tree/forest, SCC/WCC, cycles, sources/sinks/bottlenecks), BFS/DFS traversal and flow animation; and topological analysis (Boyer–Myrvold planarity, Euler characteristic, Kuratowski subgraphs, genus bound, 3D edge crossings).
• Visualization: Geometry-Nodes rendering of nodes (instanced primitives) and edges (tubes/curves), attribute-driven coloring with scientific colormaps, node and edge sizing that each read their own attribute, edge-style presets (Gephi, Cytoscape, Schematic, Bundled, Flow, Minimal), interactive gizmos/toolbar and text-overlay labels.
• SciGraphs render engine: a third engine beside Cycles and EEVEE, drawing from GPU buffers instead of instanced geometry, so it stays interactive at millions of elements. Nodes are sphere or ribbon impostors, edge styles are tessellated in GLSL, and the level of detail is chosen in pixels rather than distance. It carries the filter stack, structural simplification, hierarchical bundling, stylised shading, depth of field, a density cloud, and viewport overlays for the color key, labels, a reference grid and the node under the cursor.
• Reproducible Pipelines: declarative JSON/YAML workflow specifications with deterministic seeding and provenance manifests.
• Export: GEXF, GraphML, JSON, CSV edge lists, Pajek .net, node positions and statistics reports, plus a conversion script for PyTorch Geometric structures.

External Data Sources & APIs

SciGraphs integrates several open and commercial data services:

查看官方完整介绍 ↗