Using ‘imaginative’ AI to survey past and future earthquake damage

Using ‘imaginative’ AI to survey past and future earthquake damage

https://news.osu.edu/using-imaginative-ai-to-survey-past-and-future-earthquake-damage/

Publish Date: 2026-03-25 10:16:00

Source Domain: news.osu.edu

• Innovative AI Model: Researchers have developed an AI tool called the LoRA-Enhanced Ground-view Generation (LEGG) diffusion model to assess earthquake damage, trained on real aerial drone images to create detailed 3D reconstructions.

• 3D Ground-Level Simulations: The LEGG model generates image pairs showing building damage from above and ground-level views, which are crucial for first responders to make timely decisions.

• Filling a Research Gap: Unlike previous studies relying on aerial surveys, LEGG focuses on ground-level damage, addressing a gap in rapid assessment methods for densely populated urban areas.

• Application in Real Disasters: Researchers tested the model using the 2023 Kahramanmaras earthquake in Turkey, where it demonstrated high accuracy in identifying structural issues like cracks and collapses.

• Potential for Improved Emergency Responses: The study suggests that LEGG’s rapid and detailed simulations can offer vital information to emergency crews, enhancing rescue operations and disaster preparedness.

• Future Implications: The framework could guide the design of more resilient infrastructures, reshape post-disaster assessments, and influence emergency management policies globally.

• Researcher Insights: Rongjun Qin and his team highlight that while the model shows great potential, it requires more quality data and experiments to achieve full efficacy, aiming to predict destruction levels in other earthquake-prone regions.

• Collaborative Support: The research was collaboratively supported by various institutions and funding bodies including the Scientific and Technological Research Council of Türkiye and the Office of Naval Research.