CrossTimeEdit generates realistic historical street views across cities
CrossTimeEdit: A Decade-Spanning Cross-View Dataset and Reward-Guided Editing for Historical Street-View Generation
Computer Vision and Pattern Recognition
Summary
Urban history captured in street-view images has many gaps because some places weren't photographed often. The authors created a large dataset with images from different years and views across multiple cities to help fill these gaps. They then built CrossTimeEdit, a model that edits recent street images to show plausible past appearances by noting what has changed using satellite images. Their method improves how realistic and consistent the generated past views are compared to previous models.
What this means in practice
- •For urban planners: Generate plausible historical street views to assess urban change when photographic records are incomplete.
- •For game developers: Create realistic past cityscapes based on current street views and satellite changes for immersive historical scenes.
Authors
Hanwen Lu, Jun He, Mingjia Yang, Hao Wei, Jinhao Huang, Yi Lin, Xiang Zhang
Abstract
Historical street-view imagery records urban evolution, but uneven coverage leaves substantial gaps in historical records. Generating plausible past appearances requires restoring changed structures while preserving persistent scene content. We construct VIGOR-his, a decade-spanning cross-view dataset containing 43,653 location-level quadruplets across 11 cities on three continents. Its automated pipeline performs spatial pairing, consistency screening, change classification, and the generation and validation of satellite-based change descriptions and local editing instructions. Based on VIGOR-his, we propose CrossTimeEdit, a model that reformulates historical street-view generation as editing, using recent street views to constrain viewpoint and unchanged appearance and temporal satellite differences as change evidence. Starting from FLUX.2 [Klein] 4B, we train CrossTimeEdit through supervised fine-tuning (SFT) followed by online reinforcement learning (RL). We design three street-view editing criteria, namely Instruction Alignment (IA), Background Preservation (BP), and Quality and Physical Plausibility (QP), as both RL reward dimensions and evaluation metrics. We optimize this multi-reward objective using Within Group Relative Policy Optimization for flow-matching models (Flow-GRPO) with Group reward-Decoupled Normalization Policy Optimization (GDPO), which normalizes each reward dimension before aggregation. CrossTimeEdit improves overall performance across the three editing criteria by 17.12\% over the pretrained baseline and outperforms cross-view generation models in scene consistency, visual realism, and perceptual quality. The implementation code, dataset, and model weights are available at https://luhanwen67.github.io/CrossTimeEdit-release/.