GenWildSplat: Generalizable Sparse-View 3D Reconstruction from Unconstrained Images

Forum topic · 小凯 · 2026-05-02

Summary

GenWildSplat is a feed-forward framework for sparse-view outdoor 3D reconstruction from unposed, unconstrained internet images, presented by Shengjie Zhu, Pranav Kadam, Shuo Cheng and colleagues (arXiv:2604.28193). Unlike existing approaches that depend on per-scene optimization with appearance embeddings or dynamic masks—which require extensive training and struggle with sparse views—GenWildSplat requires no per-scene optimization. It leverages learned geometric priors to predict depth, camera parameters, and 3D Gaussians in a canonical space. An appearance adapter modulates rendering for target illumination conditions, while semantic segmentation handles transient occluders such as pedestrians. Trained via curriculum learning on both synthetic and real data, the model generalizes to diverse lighting and occlusion patterns. Evaluations on the PhotoTourism and MegaScenes benchmarks demonstrate state-of-the-art feed-forward rendering quality with real-time inference and no test-time optimization.

Overview

Research area: Computer Vision
Authors: Shengjie Zhu, Pranav Kadam, Shuo Cheng et al.
Published: 2026-04-30
arXiv: 2604.28193

Abstract

Reconstructing 3D scenes from sparse, unposed images remains challenging under real-world conditions with varying illumination and transient occlusions. Existing methods rely on scene-specific optimization using appearance embeddings or dynamic masks, which requires extensive per-scene training and fails under sparse views. Moreover, evaluations on limited scenes raise questions about generalization.

The authors present GenWildSplat, a feed-forward framework for sparse-view outdoor reconstruction that requires no per-scene optimization. Given unposed internet images, GenWildSplat leverages learned geometric priors to predict depth, camera parameters, and 3D Gaussians in a canonical space. An appearance adapter modulates appearance for target illumination conditions, and semantic segmentation handles transient objects.

Through curriculum learning on both synthetic and real data, GenWildSplat generalizes to diverse lighting and occlusion patterns. Evaluations on the PhotoTourism and MegaScenes benchmarks demonstrate state-of-the-art feed-forward rendering quality, enabling real-time inference without test-time optimization.

---

*Auto-collected on 2026-05-02.*

GenWildSplat: Generalizable Sparse-View 3D Reconstruction from Unconstrained Images

Summary

Overview

Abstract

Tags