IEEE Geoscience and Remote Sensing Magazine - March 2021 - 287

TECHNICAL COMMITTEES
NAOTO YOKOYA, PEDRAM GHAMISI, RONNY HÄNSCH, COLIN PRIEUR, HANA MALHA,
JOCELYN CHANUSSOT, CALEB ROBINSON, KOLYA MALKIN, AND NEBOJSA JOJIC

2021 Data Fusion Contest: Geospatial
Artificial Intelligence for Social Good

he 2021 Data Fusion Contest, organized by the
Image Analysis and Data Fusion Technical Committee (IADF TC) of the IEEE Geoscience and Remote
Sensing Society (GRSS), promotes research on geospatial artificial intelligence (AI) for social good. The
global objective is to build models for understanding
the state of and changes in artificial and natural environments from remote sensing data toward sustainable development. The contest is designed as a benchmark competition, following previous editions [1]-[5].
The 2021 Data Fusion Contest (Figure 1) consists of
two parallel tracks:
1) detection of settlements without electricity (DSE)
2) multitemporal semantic change detection (MSD).
Track DSE, co-organized by Hewlett Packard Enterprise, SolarAid, and Data Science Experts, addresses
the automatic detection of human settlements, deprived of access to electricity, using multimodal, multiresolution, and multitemporal satellite remote sensing data. Sentinel-1 synthetic aperture radar data and
Sentinel-2, Landsat-8, and Suomi Visible Infrared Imaging Radiometer Suite (VIIRS) nighttime images are
used as input (see Figure 2). While the original ground
sampling distance (GSD) ranges from 10 to 750 m,
all images were resampled at 10 m. Semantic labels of
four classes (i.e., settlements with and without electricity and no settlements with and without electricity) are provided at a GSD of 500 m for the training
set. Participants are required to submit binary detection maps of settlements without electricity at a GSD
of 500 m. The classification accuracy is evaluated using the F1 score. The main challenge of Track DSE is
to develop robust and efficient methods to extract
high-level semantic information from such heterogeneous data.
Digital Object Identifier 10.1109/MGRS.2021.3055633
Date of current version: 8 March 2021

MARCH 2021

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

Track MSD, co-organized by Microsoft AI for Earth,
focuses on automatic land cover change detection and
classification from multitemporal, multiresolution, and
multispectral imagery. The task of Track MSD is to create
bitemporal, high-resolution land cover maps using only
low-resolution and noisy land cover labels for training.
Participants receive a data set covering the U.S. state of Maryland,
composed of 1) 1-m multispectral
THE MAIN FEATURE OF
aerial imagery for 2013 and 2017
THE 2021 DATA FUSION
from the United States Department
CONTEST IS THAT IT IS
of Agriculture's National Agriculture Imagery Program (NAIP) data;
DIRECTLY CONNECTED TO
2) 30-m multispectral satellite imSOCIAL CONCERNS, SUCH
agery (Landsat-8) for five points in
AS ENERGY EQUALITY.
time between 2013 and 2017; and
3) 30-m noisy, low-resolution land
cover labels for 2013 and 2016
from the United States Geological Survey's National
Land Cover Database (NLCD) data (see Figure 3). Participants are required to infer high-resolution (1-m
GSD) land cover maps that identify changes between
2013 and 2017. Performance is assessed using the
intersection-over-union metric averaged across eight
287

IEEE Geoscience and Remote Sensing Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2021