CrossTimeEdit uses reward-guided editing for historical street-view generation
September 28, 2026
The CrossTimeEdit model utilizes the VIGOR-his dataset, containing 43,653 location-level quadruplets, to generate historical street-view imagery. It treats historical reconstruction as an editing task, using recent views and satellite data to constrain viewpoint and temporal consistency.
HOW THIS AFFECTS YOU
●
builderThis provides a method for generating historically accurate urban visual content.
●
researcherYou can leverage the VIGOR-his dataset for research in temporal image consistency.