EGGS: Exchangeable 2D/3DGaussian Splatting for Geometry-Appearance Balanced Novel View Synthesis

Novel view synthesis (NVS) is crucial in computer vision and graphics, with wide applications in AR, VR, and autonomous driving. While 3DGaussian Splatting (3DGS) enables real-time rendering with high appearance fidelity, it suffers from multi-view inconsistencies, limiting geometric accuracy. In contrast, 2DGaussian T Splatting o address (2DGS) these limitations, enforces multi-vie we propose w consistenc Exchangeable y but compromises Gaussian Splatting texture (EGGS), details. a hybrid representation that integrates 2D and 3DGaussians to balance appearance and geometry. To achieve this, we introduce Hybrid Gaussian Rasterization for unified rendering, Adaptive Type Exchange for dynamic adaptation between 2D and 3DGaussians, and Frequency-Decoupled Optimization that effectively exploits the implementation strengths of ensures each type efficient of Gaussian training representation.