Ask or search anything...

Researchers from the University of Hong Kong, Beijing Institute of Technology, and the University of Delaware developed UniTS, a unified time series generative model based on flow matching, capable of performing four distinct remote sensing tasks: reconstruction, cloud removal, semantic change detection, and forecasting. UniTS demonstrated superior performance across these tasks, achieving over 1.88 dB PSNR improvement for cloud removal on a new challenging dataset with 84.02% average cloud coverage, and the best mIoU scores for semantic change detection.