SENSE: Satellite-based Energy Synthesis for Sustainable Environment

Researchers introduced SENSE, a diffusion-based framework for urban building energy modeling that synthesizes aligned satellite imagery, energy maps, and height annotations.

Using road networks and urban density signals, SENSE generated high-fidelity, physically consistent city data across New York City, Boston, Lyon, and Busan.

The model worked with less than 20% labeled data and still improved prediction accuracy while reducing error versus existing methods.

By creating realistic synthetic labels where real urban energy data are scarce, SENSE can better support downstream prediction and sustainable city planning.