WildGaussians: 3D Gaussian Splatting in the Wild

While the field of 3D scene reconstruction is dominated by NeRFs due to theirphotorealistic quality, 3D Gaussian Splatting (3DGS) has recently emerged,offering similar quality with real-time rendering speeds. However, both methodsprimarily excel with well-controlled 3D scenes, while in-the-wild data -characterized by occlusions, dynamic objects, and varying illumination -remains challenging. NeRFs can adapt to such conditions easily throughper-image embedding vectors, but 3DGS struggles due to its explicitrepresentation and lack of shared parameters. To address this, we introduceWildGaussians, a novel approach to handle occlusions and appearance changeswith 3DGS. By leveraging robust DINO features and integrating an appearancemodeling module within 3DGS, our method achieves state-of-the-art results. Wedemonstrate that WildGaussians matches the real-time rendering speed of 3DGSwhile surpassing both 3DGS and NeRF baselines in handling in-the-wild data, allwithin a simple architectural framework.

Further reading#

• Access Paper in arXiv.org