BRIGHT

BRIGHT is the first open-access, globally distributed, event-diverse multimodal dataset specifically curated to support AI-based disaster response. It covers five types of natural disasters and two types of man-made disasters across 14 disaster events in 23 regions worldwide, with a particular focus on developing countries.

It supports not only the development of supervised deep models, but also the testing of their performance on cross-event transfer setup, as well as unsupervised domain adaptation, semi-supervised learning, unsupervised change detection, and unsupervised image matching methods in multimodal and disaster scenarios.