Mapping Land-Use, Hazard Vulnerability and Habitat Suitability Using Deep Neural Networks
Summary
Deep learning is a computer analysis technique inspired by the human brain’s ability to learn. It involves several layers of artificial neural networks to learn and subsequently recognize patterns in data, forming the basis of many state-of-the-art applications from self-driving cars to drug discovery and cancer detection. Deep neural networks are capable of learning many levels of abstraction, and thus outperform many other types of automated classification algorithms. This project developed software tools, resources, and two training workshops that will allow USGS scientists to apply deep learning to remotely sensed imagery and to better understand natural hazards and habitats across the Nation. The tools and training resources are [...]
Summary
Deep learning is a computer analysis technique inspired by the human brain’s ability to learn. It involves several layers of artificial neural networks to learn and subsequently recognize patterns in data, forming the basis of many state-of-the-art applications from self-driving cars to drug discovery and cancer detection. Deep neural networks are capable of learning many levels of abstraction, and thus outperform many other types of automated classification algorithms. This project developed software tools, resources, and two training workshops that will allow USGS scientists to apply deep learning to remotely sensed imagery and to better understand natural hazards and habitats across the Nation. The tools and training resources are available from GitHub links, and several publications have come from their usage.
Principal Investigator : Jonathan Warrick Co-Investigator : Daniel D Buscombe, Christopher R Sherwood, Paul E Grams, Jenna A Brown
