- data
- buildings.json - geojson, raw OSM data with 2d polygons
- reestr.csv - information about buildings
- gecoded_data.csv - info with coordinates
- result.json - geojson, OSM data joined with info
- kepler_map.html - final kepler.gl map
- notebooks
- 1.Geocoder.ipnb - example of usage Yandex.Geocoder
- 2.Matcher.ipynb - joining gecoded_data with polygons using GeoPadnas
- env.txt - installation of GeoPandas in separate python env
Now I'm going to show you how I could create a 3D map of my hometown using 2D polygons, table with info about buildings, Geocoder, GeoPandas, Kepler.gl.
Map like that
What was my task about? When i worked at Niznhy Novgorod's company 2 years ago, i got a task to create a visualization of old and emergency buildings. I got the table, there were address, number of floors, year of construction and boolean emergency flag.
| address | floor_count_max | built_year | is_alarm |
|---|---|---|---|
| д. г Навашино ул Соболева д.1 | 2 | 1954 | Нет |
| обл. Нижегородская, р-н. Ардатовский, рп. Ардатов, ул. 1 Мая, д. 32 | 5 | 1977 | Нет |
My first idea was: "Okay, i could receive coordinates and display them on the map." When I worked at Yandex.Map and we often used Yandex.Geocoder. I remembered that service well, it was our bottleneck. But it has strong terms of use. Of course, for Russia, Yandex is better, but I need to store several thousand points, so I decided to use free services. This is a good post about geocoding. I got a good idea from the comments. I can use the free service and send some problematic cases to Yandex.
Okay, i had coordinates. What's next? I discovered a perfect geo visualization tool from Uber - Kepler.gl. I made my first attempt. It was awful. There were octagons or rectangles instead of real buildings. But I can regularize their colors and height depending on different columns. Height depend on floors and colors depend on, for example, built year. I really liked it. But my boss didn't.
I started to find ways to make it better. And I found 2D polygons. I downloaded it from a site like that. Cut my region off. This is a good post about it. And i had a problem. I should have matched my coordinates of buildings with polygons. I started to write my own code based on this algorithm. As a result, it was very slow. But i did it. I got an acceptable result, and my boss was pleased.
When I decided to make a repo, I found this awesome post about the same problems, but it used GeoPandas. It's beautiful. I got a speed gain of a hundred times. It matched all coordinates with polygons for 2 seconds. You can run it.
And I get a good result. I was really proud of myself. I created my own Yandex.Map or 2GIS. And everyone can add additional information to each building, for example, images, historical information, etc. Each building has its own info card. This is a growth point.
As you can see, I had many polygons without matching. I can improve my geocoding or download more complete raw data. But for me, it was a study task, and I worked with given data.