This page provides a summary of the labeled classes and the cities involved in the Housing Passport project.

Thus far, we have worked with a total of 7 classes. However, the selection of classes for annotation has been contingent upon the unique characteristics of each city in which our work has been conducted.

The tables below provide a breakdown of building property and parts classes, along with the cities where the labeling has been done.

Additionally, there is a piece of information about the classes for rooftop mapping.

Building properties

Building parts

Labeling adjustments by city

1. For LIMA, it is the unique city where there was annotated the vintage building property
2. For PADANG, there were some adjusments on the clases:
   • We annotated only building properties: Building completeness, building material, building use, building condition.
   • We added an additional property called Building type.
   • In the Building use property the new classification was: Residential, mixed, commercial and critical infrastructure.
3. For ARMENIA, we ran inference using the existing models for both building parts and building properties that were trained with imagery from Peru and Colombia.
4. For DOMINICA, we annotated the same building properties as we did for PADANG having two exeptions:
   • We have not considered the building type class.
   • In the building material class, we have not considered the plant_material class since it was not relevant for this city.

Rooftops mapping

The rooftops mapping was done in Independencia District in Lima as a continuation of the Housing Passports work. The mapping was done using the Java OpenStreetMap tool and the properties considered on mapping were: