Google Open Buildings 2.5D Temporal

✦ DISCOVERED BY ORIONAIGoogle ResearchBuildings and infrastructureGlobal
Loading the data…Global

About this dataset

Source: https://sites.research.google/gr/open-buildings/temporal/ The Open Buildings 2.5D Temporal Dataset contains annual data spanning eight years (2016-2023) with building presence, fractional building counts, and building heights covering approximately 58 million square kilometers. This dataset requires some knowledge with using scripts. The ZIP contains .txt files for over 130 countries and territories. The primary purpose of the data is to support comparison of building footprints across multiple years.

Documentation

A deep learning model was trained to detect per pixel building presence, fractional counts and height using Copernicus Sentinel-2 imagery. We then ran inference for this model on stacks of 32 Sentinel-2 images centered around the 30th of June of each year (2016 - 2023) for most of the Global South. Only opaque-cloud-free images were used (according to the [QA60 band](https://sentiwiki.copernicus.eu/web/s2-processing)). Heights for non-building objects were masked out using the building presence output of the model. The accompanying [technical report](https://arxiv.org/abs/2310.11622) describes the training and evaluation of the model.

Related datasets

✦ Suggested by OrionAI
Google Open Buildings v3 (building footprints)

A dataset of 1.8 billion building detections across Africa, South Asia, South-East Asia, Latin America and the Caribbean, derived from high-resolution satellite imagery with a machine-learning model. Each building has its footprint polygon, area and a confidence score.

Buildings and infrastructure · Google Research
ViewSave
Building footprints: Africa (Microsoft)

AI-derived building footprint polygons for Africa from Microsoft's Global ML Building Footprints, in 219 quadkey tiles of line-delimited GeoJSON, uploaded 2026-02-23.

Buildings and infrastructure · Microsoft
ViewSave
Building footprints: Canada (Microsoft)

AI-derived building footprint polygons for Canada from Microsoft's Global ML Building Footprints, in 2055 quadkey tiles of line-delimited GeoJSON, uploaded 2026-02-23.

Buildings and infrastructure · Microsoft
ViewSave
Reunion Island: Building Footprints in Areas Impacted by Cyclone Belal

A dataset of building footprints of the Reunion Island, in the area of the January 2023 Belal Cyclone. Building footprints are useful for a range of important applications, from population estimation, urban planning and humanitarian response, to environmental and climate science. This large-scale open dataset contains the outlines of buildings derived from high-resolution satellite imagery in order to support these types of uses. For more info visit open buildings FAQ https://sites.research.google/open-buildings/#faq

Buildings and infrastructure · Google Research
ViewSave
Building footprints: Asia (Microsoft)

AI-derived building footprint polygons for Asia from Microsoft's Global ML Building Footprints, in 749 quadkey tiles of line-delimited GeoJSON, uploaded 2026-02-23.

Buildings and infrastructure · Microsoft
ViewSave
Building footprints: Brazil (Microsoft)

AI-derived building footprint polygons for Brazil from Microsoft's Global ML Building Footprints, in 1154 quadkey tiles of line-delimited GeoJSON, uploaded 2026-02-23.

Buildings and infrastructure · Microsoft
ViewSave

Indexed by OrionGIS from Humanitarian Data Exchange (HDX). Data is loaded from the publisher's own server when you use it. Permanent link