OSMnx Street Networks — Core Concepts

Learn to download, model, route, and visualize street networks from OpenStreetMap using OSMnx and NetworkX in Python.

OSMnx is a Python library that downloads geospatial data from OpenStreetMap and models street networks as NetworkX graphs. It handles the entire pipeline — query, download, clean, project, analyze, and visualize — in a few lines of code.

Downloading a network

import osmnx as ox

# By place name
G = ox.graph_from_place("Manhattan, New York, USA", network_type="drive")

# By point and radius (meters)
G = ox.graph_from_point((40.748, -73.985), dist=1000, network_type="walk")

# By bounding box
G = ox.graph_from_bbox(40.80, 40.70, -73.90, -74.00, network_type="bike")

The network_type parameter filters edges: "drive" excludes footpaths, "walk" includes pedestrian paths, "bike" includes cycleways, and "all" keeps everything.

Graph structure

The result is a MultiDiGraph — a directed graph that allows multiple edges between nodes (for one-way streets, parallel roads):

print(G.number_of_nodes())  # intersections + dead-ends
print(G.number_of_edges())  # road segments

# Node attributes
node_data = G.nodes[42431337]
# {'y': 40.748, 'x': -73.985, 'street_count': 4}

# Edge attributes
edge_data = G.edges[(u, v, 0)]
# {'length': 120.5, 'name': 'Broadway', 'highway': 'primary', 'maxspeed': '25 mph'}

Shortest-path routing

orig = ox.nearest_nodes(G, -73.990, 40.750)
dest = ox.nearest_nodes(G, -73.970, 40.730)

route = ox.shortest_path(G, orig, dest, weight="length")
# Returns a list of node IDs

# Visualize
ox.plot_graph_route(G, route, route_linewidth=3, node_size=0)

You can weight routes by "length" (shortest distance), "travel_time" (fastest with speed limits), or any custom attribute.

Adding travel time

G = ox.add_edge_speeds(G)        # infer from maxspeed tags
G = ox.add_edge_travel_times(G)  # length / speed

fast_route = ox.shortest_path(G, orig, dest, weight="travel_time")

Network statistics

OSMnx computes standard graph metrics for urban analysis:

stats = ox.basic_stats(G)
print(stats["street_density_km"])     # km of road per sq km
print(stats["intersection_density_km"])  # intersections per sq km
print(stats["circuity_avg"])          # avg route circuity (1.0 = straight line)

Higher intersection density and lower circuity indicate more walkable, grid-like neighborhoods.

Visualization

# Basic plot
fig, ax = ox.plot_graph(G, node_size=5, edge_linewidth=0.5)

# Color edges by type
ec = ox.plot.get_edge_colors_by_attr(G, "length", cmap="plasma")
fig, ax = ox.plot_graph(G, edge_color=ec, node_size=0)

OSMnx also exports to GeoDataFrames for integration with GeoPandas plotting or web maps.

Beyond streets: buildings and amenities

OSMnx downloads more than roads:

# Building footprints
buildings = ox.features_from_place("Montmartre, Paris", tags={"building": True})

# Parks
parks = ox.features_from_place("Brooklyn, NY", tags={"leisure": "park"})

# Cafés
cafes = ox.features_from_point((48.886, 2.341), tags={"amenity": "cafe"}, dist=500)

All results are GeoPandas GeoDataFrames.

Common misconception

Many people think OSMnx only works for car routing. In practice, pedestrian and bicycle networks are equally well supported, and the library’s real strength is network analysis — measuring connectivity, centrality, and structure — not just point-to-point navigation.

The one thing to remember: OSMnx converts the free OpenStreetMap database into a routable, analyzable Python graph, making street-network science accessible with a single function call.

pythonosmnxstreet-networksgeospatial