TransSplat paper introduces unbalanced semantic transport for language-driven 3DGS editing

Researchers have introduced TransSplat, a novel method for editing 3D scenes represented by Gaussian Splatting using natural language. This approach tackles the challenge of aligning semantic edits made in 2D views with the corresponding 3D elements. TransSplat formulates the editing process as an unbalanced semantic transport problem, establishing explicit correspondences between 3D Gaussians and 2D editing prototypes to guide appearance updates and improve local control precision. AI

影响 Introduces a new technique for language-driven 3D scene manipulation, potentially improving workflows in VR/AR content creation.

排序理由 This is a research paper describing a new method for 3D scene editing.

在 arXiv cs.CV 阅读 →

• 3D Gaussian Splatting
• TransSplat
• Yanhui Chen

• 论文
• 其他

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →