Abstract: Over the past decade, Oak Ridge National Laboratory (ORNL) has demonstrated robust automation, scaling, and knowledge exploitation of geospatial datasets across a diversity of geographic, temporal, cultural, and sensor conditions. This presentation will walk through an overview of GeoAI and spatio-temporal analytics tools and workflows used at ORNL. By attending this workshop, you will understand the basics of geospatial data and the tools used to analyze, process, and store geospatial data. Two hands-on examples will be presented to show how geospatial data and software are used in practice for specific applications.
Session Outline: Hands on examples and tutorials
Module 1: Geospatial data – What is geospatial data and where can I find it?
In this lesson, we give an overview of geospatial data and where you can find open-source geospatial datasets. The primary datasets we will discuss include information about demographics, infrastructure, and land-cover. We will also discuss satellite and aerial imagery data.
Module 2: Geospatial software and tools - What tools can I use to analyze, process, and store geospatial data?
We will review commonly used software and tools for geospatial data analysis, processing, and storage. This will cover desktop software, spatial databases, and packages available in the Python and R programming languages.
Module 3: Raster data application – Road segmentation from satellite imagery using deep learning. In this lesson, you will learn about open-source road segmentation benchmark datasets and a deep learning workflow to identify roads in satellite imagery.
Module 4: Vector data application – Basic concepts and considerations
In this lesson, you will be introduced to basic operations and considerations when working with non raster geospatial data. Through hands on examples, users will be introduced to helpful concepts within geocomputation and how those concepts are implemented in common open source geospatial software.
Bio: Dalton Lunga is a senior R&D scientist and research group leader for GeoAI at the Oak Ridge National Laboratory. His research interests are in domain adaptation, manifold learning, unsupervised representation learning using deep learning approaches for geospatial imagery analytics. His technical background includes image processing, statistical machine learning, remote sensing, and geospatial data analysis. He currently conducts research and development in machine learning techniques and advanced workflows for handling large volumes of geospatial data. Before ORNL, Dalton worked as machine learning research scientist at the council for scientific and industrial research in South Africa on various projects. He received his Ph.D. in electrical and computer engineering from Purdue University, West Lafayette, Indiana.