IEEE Geoscience and Remote Sensing Magazine - June 2021 - 120

◗ developing guidelines to determine how much training
data are necessary for a problem
◗ establishing requirements for compiling benchmark
training data sets and advancing model developments
◗ filling gaps in training data catalogs (based on disciplines
and application areas).
The second goal was to craft guidelines for tools and
practices related to generating, cataloging, and disseminating
training data sets to be used as benchmarks for new
model developments. The associated topics included the
following:
◗ devising best practices for disseminating training data
categorized for science disciplines
◗ building open source infrastructure to register and document
training data
◗ building tools for discovering and accessing training
data in cloud environments
◗ establishing best practices for training data selection
and model development.
Fifty-one participants from government, academia, and industry
discussed the topics over three days. They were divided
into three working groups, each of which presented
its findings at the end of the workshop.
OUTCOME
As its primary outcome, the workshop included a set of
recommendations in a written report [2]. ESDS has already
started to act on the proposals. One such activity is a data
science competition to forecast tropical cyclone intensity
[3], in which some of the workshop guidelines were implemented
to design the training data set. The authors hope
to learn from this competition and present results in a
future publication.
ACKNOWLEDGMENTS
The workshop was organized in collaboration with the
Radiant Earth Foundation. The ESI TC would like to
thank the ESDS program and Clark University for sponsoring
the workshop and the GRSS for continued support.
Furthermore, the ESI TC would like to specially thank Dr.
Rahul Ramachandran, NASA Marshall Space Flight Center,
and Dr. Hamed Alemohammad, Radiant Earth Foundation,
for helping to formulate and
organize the workshop.
Storage
Strategy
Mission
ML-Based Validation
and Anomaly Detection
Data
Automation
Lifecycle
Automated
Data Quality
AI/ML Tools for
Labeling Data
o
v
U e
s
Data-Driven Search
and Discovery,
AI-/ML-Ready Data
FIGURE 1. AI for the full data lifecycle.
120
Automated and
Consistent
Metadata
AUTHOR INFORMATION
Manil Maskey (manil.maskey@nasa
.gov) received his Ph.D. degree in computer
science from the University of
Alabama, Huntsville. He is a senior research
scientist at the NASA Marshall
Space Flight Center, Huntsville, Alabama,
35808, USA. He also leads the
advanced concepts efforts within the
Interagency Implementation and Advanced
Concepts Team. His research
interests include computer vision,
visualization, knowledge discovery,
cloud computing, and data analytics.
He chairs the IEEE Geoscience
and Remote Sensing Society Earth
Science Informatics Technical Committee
and leads machine learning
activities for the NASA Earth Science
Data Systems program. He is a Senior
Member of IEEE.
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE JUNE 2021
S
a t
s
D
d n
G
n
s
o
c
a
e
i o
e
u
m
d r
n
u t
c
e
n
r P
a
s t r
d
Archive
a
t
o
i
t
D i
a n
i
d
n
u
n
c
t
r e
i b
A c
t
o
o
I
n
g
e
s
S
i
t
o
t
User
Support
n
r
g
a
a
g
n
e
d
D
i
a
b b
i
l
s
n
c
d
r
a a
i
l
i
t
i
t
y
y
IEEE Geoscience and Remote Sensing Magazine - June 2021

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