Omar Maher, Esri; Brady Cline, Esri
We’ll present an end-to-end workflow for damage detection and disaster response using Esri’s ArcGIS AI capabilities and NVIDIA GPUs. Identifying damaged buildings and roads following disasters is a key prerequisite in allocating potentially life-saving resources. Manual identification takes significant manpower and time, which are both critical. With the increased accessibility of drone and satellite imagery, along with computer vision models, we’ve made the complete automation of damaged structure detection possible. Our workflow includes imagery access and management; model training and deployment to production; massive inference at scale; and the creation of meaningful geospatial information products.