The paper introduces Geo-Anchored Cloud Removal (GACR), a framework designed to make cloud removal useful for downstream interpretation rather than optimizing visual realism alone. Its Observation-Anchored Residual Flow (OAR-Flow) starts the generative reconstruction from the cloudy observation, while Geo-Contextual Prior Alignment (GCPA) uses a Vision Foundation Model to constrain outputs toward a semantic manifold. The authors report experiments on six cloud-removal datasets and twelve downstream tasks, claiming improved reconstruction quality and downstream accuracy. Code is available in the linked GitHub repository.
No heat snapshots are available in the last 24 hours.