Assessing building damage is a critical first step when natural disasters strike. xView2, a machine-learning (ML) computer vision system, speeds this process by using satellite imagery to classify damage to structures. In a busy 2020 wildfire season, multiple disaster response agencies received an xView2 prototype for testing.

xView2 sprang from a Defense Innovation Unit (DIU) 2018 competition that focused on accelerating progress in four computer vision frontiers: reduce minimum resolution for detection, improve learning efficiency, enable discovery of more object classes, and improve detection of fine-grained classes. To build on this initial work, xView2 contestants built ML algorithms that scanned satellite photos of disaster-struck areas and classified building damage.

An SEI team worked with the DIU team to create the challenge, the baseline ML models, a building damage scale, and xBD, a database of satellite photos with human-labeled building damage, against which competitors’ ML results would be judged. After the competition, the SEI packaged the best ML models into a prototype tool.

In 2020, the California National Guard preliminarily tested the xView2 prototype during the massive August Complex fire and smaller Valley fire. “Generally speaking, an analyst would take an entire day or two to clear a large fire area containing hundreds of structures,” said Chief Manuel Villalba, a California National Guard intelligence analyst specializing in wildfires. “According to our testing, it’s possible xView could assess it in less than 10-20 minutes.”

To learn more about xView2, visit sei.cmu.edu/our-work/projects/display.cfm?customel_datapageid_4050=295280.

Photo: California Bureau of Land Management

Photo: Raw image Copyright 2019 Maxar/DigitalGlobe. Used under a Creative Commons Attribution-NonCommercial 4.0 license (CC BY-NC 4.0). Polygons added by Carnegie Mellon University Software Engineering Institute.