CoastSight: Generative and Vision-Language AI for Disaster Intelligence
Generating realistic, controllable emergency-response scenarios from natural language — starting with state-of-the-art text-to-image synthesis, and building toward temporally coherent text-to-video simulations for training, planning, and decision support.
Our Research
Disasters like floods, hurricanes, storm surges, and wildfires are growing more frequent and severe — but there isn't nearly enough real disaster video footage to train AI on, which makes it hard to build models that can reliably read a disaster scene and support emergency response and urban planning.
CoastSight tackles this from both ends: generating realistic, AI-ready disaster video, and training vision-language models that can understand, describe, and reason about what's happening in a scene — turning raw footage into intelligence that emergency managers, planners, and coastal communities can actually use.
Our Strategy
Team members will build generative models that turn text prompts into realistic coastal-disaster video, train vision-language models to interpret scenes, summarize footage, and answer situational questions, and draw on GIS, urban-planning, and environmental data to design realistic training scenarios — then test how well these models hold up on real emergency-management and resilience tasks.
Along the way, students build hands-on skills in generative AI, vision-language modeling, computer vision, video analytics, and geospatial AI.
Our Impact
The project aims to advance next-generation AI for disaster-video understanding, close the disaster-data gap through synthetic video generation, strengthen emergency preparedness and response, and help coastal communities become safer and more resilient — while giving students hands-on experience in a fast-moving area of AI research.
Together, generative AI and video intelligence turn disaster data into insight — supporting more resilient cities and more sustainable coastal communities.
Three places to look
The project overview explains what CoastSight is and where it's headed. Meet the people behind it. The dashboard tracks the team's live action plan, timeline, and budget.
Project Overview
What CoastSight is, how the text-to-image pipeline works today, the roadmap to text-to-video generation, and the expected impact on emergency preparedness.
Explore the project → PeoplePeople
The three faculty leads and 13-student research team standing up CoastSight as a Vertically Integrated Project.
Meet the team → DashboardVIP Action Plan Dashboard
A live task tracker, recruitment & launch timeline, and funding snapshot for standing up the CoastSight VIP research team.
Open the dashboard →