Speaker: Qunying Huang, Professor of Geography, UW–Madison
The proliferation of real-time, voluminous data streams generated by diverse physical and social sensors, including satellites, unmanned aerial vehicles (UAVs), ground sensor networks, and social media, offers unprecedented opportunities for the comprehensive characterization of disaster scenarios. These data enable the development of innovative approaches to disaster mapping, damage assessment, and timely, data-driven decision-making in response to natural hazards.
Concurrently, advancements in artificial intelligence (AI), particularly deep learning (DL), have driven the evolution of geospatial AI (GeoAI) applications with the capacity to quickly and accurately extract valuable insights from extensive geospatial datasets, approximating human-like cognition. The convergence of big sensor data and GeoAI holds transformative potential for enhancing situational awareness, operational efficiencies and responsiveness in disaster management.
In light of these advancements, this talk explores key challenges, cutting-edge solutions, and real-world applications that synthesize big sensor data and GeoAI to extract actionable information for disaster response. Using wildfires and floods as case studies, I will present a suite of deep learning based models, ranging from supervised learning to self-learning and weakly supervised learning, to achieve real-time disaster mapping and damage assessment across diverse events and locations with minimal human intervention.
This seminar can also be viewed via our livestream.
Hosted by the Climate, People and the Environment Program (CPEP).