Abstract: 

Rapid developments in remote sensing technology have brought significant opportunities by providing cutting- edge optical remote sensing data in a cost-effective manner. In particular, rich spatial details embedded in high-resolution remote sensing images provide great potential to understand our environment. In this tutorial, we will cover the basic characteristics of remote sensing images, and introduce open source software and deep learning libraries for geospatial data analysis. We will present the use of a convolutional neural network architecture to perform the tasks of object detection and segmentation with remote sensing images spanning very large geographical extents. We will also talk about the multi-GPU training that enables accelerated geospatial intelligence extraction.​

Bio: 

Dalton D. Lunga is a geospatial and machine learning scientist at the Oak Ridge National Laboratory. He has extensive experience on designing and developing computational methods for knowledge discovery from large volumes of high dimensional image and geospatial data. He currently leads research on deep learning feature extraction for domain adaptation and active learning, manifold learning with scalable deep learning approaches for World scale human settlement mapping. His technical background includes image processing, statistical machine learning, geospatial data analysis, visual analytics and natural language processing. He received his PhD in Electrical and Computer Engineering from Purdue University, West Lafayette, Indiana.