How to map yurts in Ulaanbaatar and save time using open source AI tools

Why yurts?

Ulaanbaatar (,47.937303861545516,19/square/1302310/59e62b7a3d6412ef7220952e?_k=nkhprp)
Ulaanbaatar (
Ulaanbaatar city area

Existing solutions

Phase 1. Let’s make simple model and try to use it!


All squares 2.5x2.5km (EPSG:3857)



QGIS polygons to centroids to circles processing
Before — red, after — blue
Inference in rural areas
False positives
Green — phase 1 territory, blue — phase 2 territory
  • 35k true positives
  • 65k false positives
  • 2k false negatives
  • precision 0.35
  • recall 0.94
  • f1 0.51

Phase 2. Let’s remove false positives and do best!

Red — positive chips
Chip squares
Phase 1 model with Phase 2 chips
Phase 1 model with yurts within Phase 2 chips


  • 62k true positives
  • 11k false positives
  • 6k false negatives
  • precision 0.84
  • recall 0.91
  • f1 0.87

Phase 3. Let’s merge it together!

Merged result


  • AI assisted mapping 12+1+2+1+4+24+2+2+1+4+16+8=77 hours
  • Mapping by hands 100k items / 0.16k items per hour = 625 hours




ML engineer

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Uplink channel mappings

AI Classification. Antique Ring Classifier

Udacity Self-Driving Car Nanodegree Project 12 — Semantic Segmentation

What is Inductive Bias in ML?

Support Vector Regression

Less than a simple introduction to Neural Networks (4/4)

Logistic Regression— Machine Learning Algorithms with Implementation in Python

Can Machine Learning predict Poverty?

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


ML engineer

More from Medium

Using Sentinel-1 SAR data for Imagery Intelligence — Detection of vessels

Batch Uploading Rasters to Earth Engine

Serving up SpaceNet Imagery for Bokeh

HIECTOR: Hierarchical object detector at scale