SeasonScapes: Learning Large-scale Re-lightable 3D Landscapes with Seasonal Variation from Sparse Webcams
Enables scalable 3D reconstruction of outdoor scenes with seasonal dynamics for computer graphics and remote sensing, though limited to sparse-view webcam data.
SeasonScapes reconstructs large-scale (50 km x 60 km) 3D landscapes with seasonal appearance variation from over 85,000 sparse webcam images, enabling relighting via physically-based rendering after diffusion-based inpainting.
We introduce SeasonScapes framework and a the SeasonScapes dataset: Swiss Sparse-view Mountain Scenes with Seasonal Changes that covers over 50 km x 60 km, composed of more than 85,000 webcam images captured from 32 different locations across 13 timestamps throughout a full year. By projecting these timestamp-specific images onto a 3D mesh, we construct seasonal 3D landscapes that reflect natural appearance changes over time. To address occlusions and missing data, we leverage conditional diffusion models for image-guided inpainting directly on the mesh. The resulting completed meshes can be further relighted using standard physically-based renderer.