Selective Style Gaussian: Enabling Localized Style Transfer in 3D Scenes

In recent years, research on 3D reconstruction techniques has intensified, particularly those involving Gaussian Splatting. Leveraging these advancements, various methodologies have been developed to edit 3D scenes. The 3D StyleGaussian model has been recognized as an innovative approach, facilitating rapid style transfers with notable efficacy via 3D Gaussian Splatting. Despite its advantages, the model is limited by a uniform style application across the scene, which restricts editing flexibility and versatility. To overcome this challenge, we introduce the Selective Style Gaussian, a novel method combining mask generation with positional matching. This technique enables targeted style applications, refining the framework and broadening the utility of 3D style transfers. The Selective Style Gaussian is expected to substantially improve the creativity and user experience in 3D editing applications by providing greater control and adaptability.