DySENSE generates synthetic city energy maps from satellite imagery
Researchers behind the arXiv paper DySENSE describe a generative AI framework that pairs static urban satellite imagery with dynamic monthly building energy consumption maps, reporting gains in downstream prediction across four cities.
In plain words
The DySENSE team says it built a generative AI system that creates synthetic satellite-style images of cities together with matching monthly building energy maps. The aim is to produce labelled energy data for cities that lack it, which the authors say could support energy planning. Results come from the authors' own experiments in four cities.
What was announced
The authors frame the work against a data gap: they write that existing studies are largely predictive for existing cities rather than generative for future ones, and that fine-grained, aligned monthly energy data are scarce. DySENSE separates long-term static urban morphology from short-term dynamic energy consumption. Conditioned on static urban land use, it generates new urban imagery with aligned energy maps; conditioned on monthly and climate prompts, it generates monthly energy maps while keeping the urban form fixed.
In experiments across New York, Washington, Norfolk and Busan, the authors report that DySENSE produced annotated synthetic energy data using less than 25% labelled data, raised downstream prediction performance by 8% mIoU, and improved prediction accuracy by 8.7% while reducing error by 9%-18% CVRMSE under the ASHRAE measure, compared with existing methods. The paper links to a dataset and code repository, but the abstract alone does not establish independent evaluation or how the framework performs outside these four cities.
What it does not show
- Evidence is limited to the paper's abstract; the full methodology, experimental setup and peer-review status are not covered, and the reported figures are the authors' own, not independently confirmed.
For security and IT teams
- Scope
- The reported results cover four cities: New York, Washington, Norfolk and Busan.
Sources
Source-support review completed 17 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.
Coverage: arXiv: abstract.
Updates and corrections
No corrections so far.
First published 17 Nov 2026
New to the subject? AI from the Ground Up starts from the beginning.