Landmarks (OSM Data)#

Four scores are here computed to extract computation landmarks: structural, visual, pragmatic. Currently, the user has to provide the building shapefile, collected independently. The height is the only necessary field, plus two datasets containing respectively information about cultural/historical landmarks (e.g. position of listed important buildings) and land-use.

[ ]:

import osmnx as ox, matplotlib.cm as cm, pandas as pd, numpy as np, geopandas as gpd
import matplotlib.pyplot as plt
from importlib import reload
import functools

pd.set_option('display.float_format', lambda x: '%.3f' % x)
pd.options.mode.chained_assignment = None

import cityImage as ci

Functions from the cityImage API used in this notebook#

This notebook scores buildings as candidate landmarks, combining structural, visual, cultural and pragmatic salience into global and local imageability scores, starting from OpenStreetMap data.

  • buildings_from_osm — Download building footprints from OpenStreetMap into the cityImage buildings schema.

  • network_from_osm — Download a street network from OpenStreetMap and return it as cityImage nodes and edges GeoDataFrames.

  • structural_score — Compute the structural landmark component (distance from the network, 2D advance visibility, number of neighbours).

  • visibility_score — Compute the visibility landmark component (facade area and, when available, 3D sight-line visibility).

  • cultural_score — Compute the cultural landmark component from historical/cultural attributes.

  • classify_land_use — Classify raw land-use values into cityImage land-use categories.

  • pragmatic_score — Compute the pragmatic landmark component from land-use frequency.

  • compute_global_scores — Combine the component scores into a global (city-wide) landmark score.

  • compute_local_scores — Combine the component scores into a local (neighbourhood-relative) landmark score.

  • plot_grid_gdfs_column — Plot the same column across several GeoDataFrames as a grid of maps.

  • rand_cmap — Build a random categorical colormap (useful for districts/partitions).

For the full list of functions and their parameters, see the API reference.

Obtaining the Data#

Fill the list ‘places’ with the names of the cities of interest. Buildings footprints are dowloaded for each city from OpenStreetMap within a certain distance from the centre of each city. The distance is set by the user. Each geodataframe is projected automatically (check OSMNx documentation for more information). The geodataframes are then cleaned and only relevant columns are kept.

[ ]:

distance_from_center = 1000
places = ['London, UK', 'Torino, Italy', 'Berlin, Germany', 'Beograd, Serbia', 'New York, US', 'Cairo, Egypt']
buildings_gdfs = []

for place in places:
    buildings_gdf = ci.buildings_from_osm(place, download_method="distance_from_address", crs=None, distance=distance_from_center)
    buildings_gdfs.append(buildings_gdf)

Visualisation. The cell below maps the result using the cityImage plotting helpers such as plot_gdf; see the Plotting section of the API reference for all styling options.

[ ]:

fig = ci.plot_grid_gdfs_column(gdfs = buildings_gdfs, titles = places, figsize = (15,15), color = 'grey', black_background = True,
                              nrows = 2, ncols = 3)

Structural, Visual and Cultural score computations#

Structural Component#

It considers:

  • The area of the building.

  • Its distance from the road.

  • The number of adjacent buildings, within the given neighbours_radius, around it.

  • A measure of 2d advance visibility, expanded till a maximum distance from the center of the building on the basis of the advance_vis_expansion_distance parameter.

[ ]:

edges_graphs = []
for n, city in enumerate(places):
    # reading the street network
    _, edges_graph = ci.network_from_osm(city, download_method="distance_from_address", network_type="walk", crs=buildings_gdfs[n].crs, distance=distance_from_center)
    ox.projection.project_gdf(edges_graph)
    edges_graphs.append(edges_graph)

for n, city in enumerate(places):
    # structural score
    buildings = buildings_gdfs[n]
    obstructions = buildings.copy()
    edges_graph = edges_graphs[n]
    buildings_gdfs[n] = ci.structural_score(buildings, obstructions, edges_graph, advance_vis_expansion_distance = 300,
                                neighbours_radius = 150)

Visual Component#

[ ]:

for n, city in enumerate(places):
    # visual score: not really computed at the moment as in most of the cases height information is not provided in OSM
    buildings_gdfs[n] = ci.visibility_score(buildings_gdfs[n])

Cultural Component#

[ ]:

for n, city in enumerate(places):
    # culturalscore:
    buildings_gdfs[n] = ci.cultural_score(buildings_gdfs[n], from_OSM = True)

Pragmatic Score#

Land use#

Before computing the pragmatic score, the land_use_raw raw field is categorised into more granular categories. One can fill in the following categories to accomodate descriptors not accounted for. Essentially, the user needs to come up with land-use categorise and distribute the OSM’s ones within these categories.

[ ]:

OSM_uses = set()  # Using a set directly for uniqueness
for n, city in enumerate(places):
    OSM_uses.update(buildings_gdfs[n].land_uses_raw.unique())  # Add unique values directly to the set

OSM_uses = list(OSM_uses)  # Convert to list after loop if needed
OSM_uses

The categorisation below has been used for larger urban areas and include OSM’s land uses that are not included in the GeoDataFrames used in this analysis.

The lists are created to contain the more specific land use descriptors derived from OSM.

[ ]:

adult_entertainment = ['brothel', 'casino', 'gambling', 'nightclub', 'private_club', 'stripclub', 'swingerclub']
agriculture = ['agricultural and forestry', 'barn', 'building in the botanical garden', 'greenhouse', 'greenhouse (botany)', 'shed', 'silo', 'stable']
attractions = ['aquarium', 'attraction', 'attraction and leisure', 'attractions', 'bath house, thermal baths', 'boathouse', 'entertainment hall',
               'exhibition hall, trade hall', 'information', 'recreation and amusement park', 'sauna', 'terrace',
               'tourist information center', 'viewpoint', 'zoo']
business_services = ['administrative building', 'atm', 'bank', 'bureau_de_change', 'conference_centre', 'coworking_space', 'customs office',
                     'data_center', 'facility building', 'foundation', 'insurance', 'office', 'offices', 'post', 'post_office', 'post_office;atm',
                     'residential building with trade and services', 'service', 'stock_exchange', 'tax office', 'tax_office', 'trade and services',
                     'trade and services building']
commercial = ['bicycle_rental', 'boat_rental', 'boat_sharing', 'books', 'car_rental', 'car_sharing', 'car wash', 'car_wash', 'commercial',
              'commercial and industry buidling', 'commercial and industry building', 'commercial building',
              'commercial building to traffic facilities (general)', 'commercial land', 'commercial services', 'commercial;educa',
              'crematorium', 'dive_centre', 'driving_school', 'fuel', 'funeral parlor', 'gas station', 'internet_cafe', 'kiosk',
              'mall', 'marketplace', 'pharmacy', 'pumping station', 'retail', 'retail;yes', 'shop', 'supermarket']
culture = ['arts_centre', 'cinema', 'club_house', 'cultural facility', 'cultural_centre', 'culture', 'exhibition_centre', 'gallery',
           'music_school', 'museum', 'opera_house', 'planetarium', 'studio', 'theater', 'theatre']
eating_drinking = ['accommodation, eating and drinking', 'bar', 'bbq', 'biergarten', 'cafe', 'cafe;restaurant', 'fast_food', 'food_court',
                   'ice_cream', 'kitchen', 'pub', 'restaurant']
education_research = ['education and research', 'research', 'research_institute', 'research_institu', 'university', 'university building']
emergency_service = ['emergency_service', 'fire brigade', 'fire_station', 'police', 'ranger_station', 'resque_station', 'security']
general_education = ['childcare', 'children home', 'college', 'education', 'education and health', 'general education school', 'kindergarten',
                     'language_school', 'nursery', 'school']
hospitality = ['building for accommodation', 'guest_house', 'hostel', 'hotel', 'hotel, motel, pension', 'refuge']
industrial = ['construction', 'factory', 'gasometer', 'industrial', 'manufacture', 'manufacturing and production', 'production building', 'workshop']
medical_care = ['clinic', 'dentist', 'doctors', 'healthcare', 'hospital', 'medical care', 'mortuary', 'nursing_home', 'retirement home',
                'sanatorium, nursing home', 'veterinary']
military_detainment = ['barracks', 'bunker', 'bunker, shelter', 'general aviation', 'military', 'penitentiary', 'prison']
monument = ['arch', 'bridge', 'castle', 'gate', 'gatehouse','monument', 'palace', 'ruins', 'tower', 'wall']
other = ['abandoned', 'air_shaft', 'block', 'boat', 'canopy', 'container', 'converter', 'exempt', 'exempt 121a', 'garage',
         'garages', 'gazebo', 'houseboat', 'hut', 'no', 'parking', 'pavilion', 'picnic_site', 'roof', 'ship', 'shelter', 'steps', 'storage_tank',
         'toilet', 'toilets', 'ventilation_shaft', 'viaduct', 'warehouse']
public = ['community_centre', 'civic', 'courthouse', 'district government', 'embassy', 'government', 'library', 'public', 'public infrastructure',
          'public_building', 'residential building with public facilities', 'townhall', 'trade_union', 'yes;public']
religious = ['basilica', 'cathedral', 'chapel', 'church', 'convent', 'monastery','mosque', 'nuns home', 'place_of_worship', 'rectory',
             'religious building', 'religious buildings', 'synagogue', 'vocational school']
residential = ['apartments', 'apartments 4-6 units', 'apartments 7 units above', 'building', 'building usage mixed with living',
               'commercial condo unit', 'condominium master', 'condominium parking', 'detached', 'exempt', 'flats', 'house', 'houses',
               'mixed use res/comm', 'NaN', 'residential', 'residential building', 'residential condo unit', 'residential land',
               'residential single family', 'residential three family', 'residential two family', 'student dorm', None]
social = ['community buidling', 'community centre', 'community hall', 'commercial social facility', 'dormitory', 'recreational',
          'social centre', 'social_facility', 'social services', 'social services building', 'society']
sport = ['bowling hall', 'grandstand', 'gym', 'horse riding school', 'indoor swimming pool', 'sport and entertainment',
         'sport hall', 'sports building', 'sports hall', 'sports or exercise facility', 'stadium', 'swimming pool']
transport = ['bus_station', 'ferry_terminal', 'light rapid transit station', 'light rapid transit station and railway station',
             'london underground station', 'railway building', 'railway station', 'railway station and london underground station', 'road transport',
             'shipping facility building', 'station', 'subway_entrance', 'train_station', 'transport', 'transportation']
utilities = ['boiler house', 'electricity substation', 'electricity supply', 'electricity', 'gas supply', 'heating plant', 'smokestack',
             'supply systems', 'telecommunication', 'waste management', 'waste treatment building', 'waste water treatment plant', 'water supply',
             'water works']

Then, the user needs to match the lists representing a Land use with the corresponding labels.

[ ]:

categories = [adult_entertainment, agriculture, attractions, business_services, commercial, culture, eating_drinking,
       education_research, emergency_service, general_education, hospitality, industrial, medical_care, military_detainment, monument,
       other, public, religious, residential, social, sport, transport, utilities]
strings = ['adult_entertainment', 'agriculture', 'attractions', 'business_services', 'commercial', 'culture', 'eating_drinking',
       'education_research', 'emergency_service', 'general_education', 'monument', 'hospitality', 'industrial', 'medical_care',
           'military_detainment', 'other', 'public', 'religious', 'residential', 'social', 'sport', 'transport', 'utilities']

Finally, one can check if there are any OSM land uses that have not been assigned to a category.

[ ]:

for n, city in enumerate(places):
    all_uses = [item for sublist in categories for item in sublist]
    land_uses = list(buildings_gdfs[n].land_uses_raw.unique())
    to_categorise = [item for item in land_uses if item not in all_uses]
    print(to_categorise)

Now, the land use information can ben finally used to derive the pragmatic score.

The radius parameter indicates the extension of the area that is used to compute the score of a building, on the basis of its land-use’s frequency, in an area of ‘buffer’ meters around it.

Score computation#

[ ]:

for n, city in enumerate(places):
    buildings_gdfs[n]['land_uses_raw'] = buildings_gdfs[n]['land_uses_raw'].str.lower() # lower everything
    buildings_gdfs[n] = ci.classify_land_use(buildings_gdfs[n], raw_land_use_column='land_uses_raw', new_land_use_column='land_use', categories=categories, strings=strings) # categories land uses
    buildings_gdfs[n] = ci.pragmatic_score(buildings_gdfs[n], search_radius = 200)

Visualisation. The cell below maps the result using the cityImage plotting helpers such as plot_gdf; see the Plotting section of the API reference for all styling options.

[ ]:

nr_categories = max([len(gdf['land_use'].unique()) for gdf in buildings_gdfs])
cmap = ci.rand_cmap(nr_categories, type_color = 'bright')

[ ]:

fig = ci.plot_grid_gdfs_column(buildings_gdfs,titles = places, cmap = cmap, column = 'land_use', nrows = 2, ncols = 3,
                              legend = True, figsize = (15,10))
plt.tight_layout(pad=3.0, rect=[0, 0, 0.85, 1])  # Adjust rect to leave space on the right
plt.show()

Final scores#

Weights definition.#

Weights of components and indexes that are not computed due to data availability are set to 0#

[ ]:

# global landmarkness components weights
g_cW = {'vScore': 0.00, 'sScore' : 0.60, 'cScore': 0.10, 'pScore': 0.30}
# global landmarkness indexes weights, cScore and pScore have only 1 index each
g_iW = {'3dvis': 0.00, 'fac': 0.00, 'height': 0.00, 'area': 0.50, '2dvis': 0.00, 'neigh': 0.30 , 'road': 0.20}

# # local landmarkness components weights
l_cW = {'vScore': 0.05, 'sScore' : 0.45, 'cScore':0.10 , 'pScore': 0.45}
# # local landmarkness indexes weights, cScore and pScore have only 1 index each
l_iW = {'3dvis': 0.00, 'fac': 0.00, 'height': 0.00, 'area': 0.50, '2dvis': 0.00, 'neigh': 0.30 , 'road': 0.20}

[ ]:

# Global and Local Score (1500 mt area)
for n, city in enumerate(places):
    buildings_gdfs[n] = ci.compute_global_scores(buildings_gdfs[n], g_iW,g_cW)
    buildings_gdfs[n] = ci.compute_local_scores(buildings_gdfs[n], l_iW, l_cW, 1500)

Visualisation#

[ ]:

scheme_dict = {'bins' : [0.125, 0.25, 0.5, 0.75, 1.00], 'scheme' : 'User_Defined'}

fig = ci.plot_grid_gdfs_column(buildings_gdfs, main_title = 'Structural score', titles = places, classes = 7, column = 'sScore_sc', nrows = 2, ncols = 3,
                cmap = 'magma', legend = True, figsize = (10,10), **scheme_dict)
plt.tight_layout(pad=0.0, rect=[0, 0, 0.90, 1.05])
plt.show()

[ ]:

fig = ci.plot_grid_gdfs_column(buildings_gdfs,main_title = 'Pragmatic score', titles = places, classes = 7, column = 'pScore', nrows = 2, ncols = 3,
                cmap = 'magma', legend = True, figsize = (10,10), **scheme_dict)
plt.tight_layout(pad=0.0, rect=[0, 0, 0.90, 1.05])
plt.show()

[ ]:

fig = ci.plot_grid_gdfs_column(buildings_gdfs, main_title = 'Local score', titles = places, classes = 7, column = 'lScore_sc', nrows = 2, ncols = 3,
                cmap = 'magma', legend = True, figsize = (10,10), **scheme_dict)
plt.tight_layout(pad=0.0, rect=[0, 0, 0.90, 1.05])
plt.show()

[ ]:

fig = ci.plot_grid_gdfs_column(buildings_gdfs, main_title = 'Global score', titles = places,  column = 'gScore_sc', nrows = 2, ncols = 3,
                cmap = 'magma', legend = True, figsize = (10,10), **scheme_dict)
plt.tight_layout(pad=0.0, rect=[0, 0, 0.90, 1.05])
plt.show()