rAMP Building Footprint Training Dataset: Shanghai, China

Provider
DevGlobal (rAMP project)
License
CC-BY-SA-4.0 (source)
Stored at
Azure Blob Storage, https://radiantearth.blob.core.windows.net/mlhub/ramp_shanghai_china
Access
Free and public. No account with the data provider is needed.
DOI
10.34911/rdnt.grvh9e
Category
Earth Observation & Geospatial
Source
https://doi.org/10.34911/rdnt.grvh9e

About rAMP Building Footprint Training Dataset: Shanghai, China

A chipped training set for building footprint detection over Shanghai, China, released by DevGlobal as part of rAMP (Replicable AI for Microplanning), an open model and data effort to map buildings for public health microplanning and disaster response. Each 256 x 256 pixel tile (or smaller at scene edges) of 30 cm imagery is paired with a GeoJSON file of building polygons drawn as rooftop outlines. This is a rAMP Tier 1 set, meaning the labels were reviewed and corrected tile by tile with the Chippy Checker editor, and it was used to train the rAMP baseline U-Net model. It contains 3,574 tiles and 7,118 buildings and was derived from SpaceNet 2 satellite imagery, tiled down from 650 x 650 pixel chips with labels revised to rooftop outlines. Keywords: urban, dense settlement. Measured 2026-11-15: source/ holds the 3,574 GeoTIFF image tiles and labels/ the 3,574 matching GeoJSON label files, with a PDF documentation file at the top level. Read with rasterio and geopandas, or with the rAMP code from DevGlobal. Released under CC BY-SA 4.0; cite DevGlobal (2022), ramp Building Footprint Training Dataset - Shanghai, China, Version 1.0, Radiant MLHub, doi:10.34911/rdnt.grvh9e.

How to access rAMP Building Footprint Training Dataset: Shanghai, China in Python

import adlfs
fs = adlfs.AzureBlobFileSystem(account_name='radiantearth', anon=True)
fs.ls('mlhub/ramp_shanghai_china/')[:10]

Lists the files with anonymous access. Runs on any machine, with no cloud account.

Data structure

Files

Example file: labels/00080f3b-446f-4af7-b0dc-85f93bef947c.geojson

Analyze this dataset without downloading it

On Scigantic, rAMP Building Footprint Training Dataset: Shanghai, China opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Earth Observation & Geospatial datasets