Photo-realistic flood scenes

FloodGen is an advocacy tool that uses generative AI to create photorealistic scenes of predicted flooding that evoke a sense of urgency about risks and impacts. Users begin with a citywide view of flood-prone neighborhoods, and can zoom in to more than 30 high-risk, high-vulnerability locations in 10 case study stites. These sites were selected based on coastal and stormwater flood hazards, vulnerability of environmental justice areas and hurricane evacuation zones, and proximity to transit, public housing, hospitals, commercial areas, schools, and libraries. The images are transformed using CycleGAN, a generative adversarial network (GAN) model that reprocesses normal street view images to depict flooding scenarios.

The goal of the project is to raise awareness of flood risks and increase support for resilience efforts. FloodGen aims to help communities prepare for floods, seek resilience funding, and engage with local governments.