google buildings extract for given area.

google-buildings-extract-bounds.py

MINLAT = 21.0762
MAXLAT = 21.2057
MINLON = 78.9929
MAXLON = 79.1738 

DATA_FILE = '3bd_buildings.csv'
OUTPUT_CSV = 'nagpur.csv'

import json, os

def write2file(line):
    with open(OUTPUT_CSV, 'a') as file:
        file.write(line)

def checkBounds(lat, lon):
    global MINLAT, MAXLAT, MINLON, MAXLON
    if (MINLAT <= lat <= MAXLAT) and (MINLON <= lon <= MAXLON):
        return True
    else:
        return False

#############
print("Starting script")

counter = 0
foundCounter = 0
with open(DATA_FILE, 'r') as f:
    for line in f:

        if counter == 0:
            # write header to output file and skip
            write2file(line)
            counter += 1
            continue
        latlon = line.split(',')[:2]
        lat = float(latlon[0])
        lon = float(latlon[1])

        # Check if the point is inside the polygon
        is_inside = checkBounds(lat, lon)
        if is_inside:
            write2file(line)
            foundCounter += 1
        
        counter += 1
        
        if counter % 10**5 == 0:
            print(f"Parsed {counter/(10**6)}M lines, found {foundCounter}")

print(f"done, parsed {counter/(10**6)}M rows total, got {foundCounter/(10**3)}K matches")

google-buildings-extract-polygon.py

DATA_FILE = '39f_buildings.csv'
AREA_FILE = 'jalpaiguri.geojson'
OUTPUT_CSV = 'output.csv'

# pip install shapely

import json, os
from shapely.geometry import Point, Polygon

def write2file(line):
    with open(OUTPUT_CSV, 'a') as file:
        file.write(line)

def fetch_polygon(filename):
    with open(filename, 'r') as f:
        polygon_data = json.load(f)
        polygon_coordinates = polygon_data['features'][0]['geometry']['coordinates'][0]
        polygon = Polygon(polygon_coordinates)
    return polygon


#############

polygon = fetch_polygon(AREA_FILE)
print("polygon loaded")

counter = 0
foundCounter = 0
with open(DATA_FILE, 'r') as f:
    for line in f:

        if counter == 0:
            # write header to output file and skip
            write2file(line)
            counter += 1
            continue
        latlon = line.split(',')[:2]
        lat = float(latlon[0])
        lon = float(latlon[1])
        # Create a Point object
        point = Point(lon, lat)

        # Check if the point is inside the polygon
        is_inside = point.within(polygon)
        if is_inside:
            write2file(line)
            foundCounter += 1
        
        counter += 1
        
        if counter % 10**5 == 0:
            print(f"Parsed {counter/(10**6)}M lines, found {foundCounter}")

print(f"done, parsed {counter/(10**6)}M rows total, got {foundCounter/(10**3)}K matches")

z_README.md

Steps for extraction by lat-long bounds

1. Fetch the .gz from https://sites.research.google/open-buildings/#download
2. Unzip it
3. Get your max and min lat-long values. You can use https://geojson.io or any other tool
4. Populate them in the top lines in the -bounds python script. Take care not to put min in max etc.
5. Change the filename values as appropriate
6. Run the -bounds python script
7. You'll see the output CSV file
8. Open QGIS
9. Menu > Layer > Add Layer > Add Delimited Text Layer
10. Select the file, make sure WKT / Well Known Text option is chosen. Press Add
11. Now you see your shapes in QGIS map
12. Right-click layer, Export
13. Export to Geojson or the shapefile format of your choice

Steps for extraction by polygon/shape

1. Fetch the .gz from https://sites.research.google/open-buildings/#download
2. Unzip it
3. Draw the area you want in https://geojson.io (single shape only, modify the script if you do multi-polygon etc), and save the file as .geojson
4. pip install shapely
5. Change the filename values at top
6. Run the -polygon python script
7. You'll see the output CSV file
8. Open QGIS
9. Menu > Layer > Add Layer > Add Delimited Text Layer
10. Select the file, make sure WKT / Well Known Text option is chosen. Press Add
11. Now you see your shapes in QGIS map
12. Right-click layer, Export
13. Export to Geojson or the shapefile format of your choice

Details

• This takes the centroid lat-long which are the first two columns in the source CSV data, and does a lat-long comparison (for bounds) or point-in-polygon operation (for polygon) on every line. Dumps the matching ones into output file
• The scripts loads and processes only one line at a time, so should work on the very large CSV files on regular / entry-level laptops etc also. Let me know if that's not happening for you.

Clarifications

• All mentioned files must be located in the same folder and this script also in that same folder. Else put in folder paths as needed.