"WRIVA CVGL Challenge 2026"

"Ground-level images provide unique contextual and semantic information that complements satellite and aerial observation. Establishing reliable connections between these modalities - addressing challenges due to near-orthogonal viewpoint differences, dramatic scale and resolution disparities, and occlusion effects - unlocks a broad range of emerging applications. These include fine-grained geo-referencing of unlocalized photos, satellite data attribution, 3D scene reconstruction across varying altitudes, multi-view data fusion for digital twins, and cross-source change detection for environmental and infrastructure monitoring.Recent advances in deep learning, cross-view retrieval, and multi-modal representation learning have markedly improved our ability to associate ground images with overhead observations; however, the task remains extremely challenging: differences in geometry, radiometry, and content visibility persist, and globally accurate ground-to-satellite localization is still limited to tens of meters without additional priors. The next frontier lies in local-context geo-localization, where approximate camera positions (e.g., within a few city blocks) enable meter-level alignment and open opportunities for downstream mapping, navigation, and situational awareness.This competition seeks to promote innovation and reproducibility for local-context ground-to-satellite image localization. Participants will receive sets of one or more satellite images and one or more ground-level images with approximate locations (e.g., within a few hundred meters) and will be evaluated on predicted camera geolocation accuracy in meters. Winning teams will be acknowledged during a session at IGARSS 2026 and invited to contribute to an article on the competition to be published in a special issue of Photogrammetric Engineering & Remote Sensing (PE&RS)."