SeasonScapes: Learning Large-scale Re-lightable 3D Landscapes with Seasonal Variation from Sparse Webcams

2026-05-09

AI Summary

This study addresses the challenge of reconstructing large-scale, relightable 3D landscapes that capture seasonal variations from sparse and unstructured webcam imagery. To this end, we propose a novel framework that integrates 3D mesh projection with image-guided conditional diffusion inpainting to effectively handle occlusions and missing data, complemented by physically based rendering for photorealistic appearance reconstruction. Our key contribution is the creation of SeasonScapes—the first dataset spanning a 50 km × 60 km region and comprising over 85,000 multi-temporal images—alongside a relightable 3D landscape generation pipeline. This work achieves, for the first time, high-quality, large-scale 3D scene reconstruction with realistic seasonal dynamics.

Abstract

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.

Problem

• 3D landscapes
• seasonal variation
• sparse webcams
• relighting
• occlusions

Innovation

• Seasonal 3D Reconstruction
• Conditional Diffusion Inpainting
• Relightable 3D Landscapes
• Sparse Webcam Imagery
• Large-scale Scene Modeling