AI tool turns drone images into street-level views to speed up earthquake response assessment

Researchers have developed a new artificial intelligence (AI) tool designed to support rapid post-earthquake damage assessment, according to recent research findings. The system aims to help emergency response teams quickly identify severely affected areas, reducing reliance on time-consuming manual inspections.

The tool, known as LEGG, uses drone imagery to generate realistic street-level views of buildings. This capability is significant, as aerial images alone do not fully reflect the perspective available to responders operating at ground level.

The AI model is trained using paired datasets that link aerial and ground-level views of the same structures, enabling it to learn the relationship between the two perspectives. Previous approaches have primarily relied on aerial imagery, which can limit the ability to interpret structural damage as it appears at close range. In contrast, conventional field assessments may require several days to complete, creating delays in emergency response.

The system was tested using data from the 2023 earthquake in Türkiye, where it demonstrated the ability to identify key damage features such as cracking, structural deformation, and partial collapse. Notably, the model achieved these results despite being trained on a relatively limited dataset.

Its performance is attributed to the integration of multiple forms of visual data, allowing the model to capture a more comprehensive representation of structural damage. The researchers suggest that, with expanded datasets and further development, such AI tools could support not only disaster response but also long-term urban planning and resilience strategies.

However, the system is intended to complement rather than replace existing assessment methods. Further validation and refinement are required before large-scale deployment, but the approach demonstrates the potential of AI to enhance decision-making in time-critical emergency situations.