View Consistent Editing

View-consistent editing aims to modify 3D scenes from multiple viewpoints while maintaining visual coherence across all perspectives. Current research heavily utilizes 3D Gaussian splatting as a scene representation, often incorporating 2D diffusion models for image editing and employing techniques like cross-attention and iterative refinement to ensure consistency between edited views. This focus on multi-view consistency is crucial for improving the realism and quality of 3D scene manipulation in applications ranging from virtual reality to computer-aided design. The development of more robust and efficient methods for view-consistent editing is driving advancements in both computer vision and computer graphics.