EGGS: Exchangeable 2D/3D Gaussian 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 3D Gaussian Splatting (3DGS) enables real-time rendering with high appearance fidelity, it suffers from multi-view inconsistencies, limiting geometric accuracy. In contrast, 2D Gaussian Splatting (2DGS) enforces multi-view consistency but compromises texture details. To address these limitations, we propose Exchangeable Gaussian Splatting (EGGS), a hybrid representation that integrates 2D and 3D Gaussians 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 3D Gaussians, and Frequency-Decoupled Optimization that effectively exploits the strengths of each type of Gaussian representation. Our CUDA-accelerated implementation ensures efficient training and inference. Extensive experiments demonstrate that EGGS outperforms existing methods in rendering quality, geometric accuracy, and efficiency, providing a practical solution for high-quality NVS.

EGGS achieves the best balance between geometric accuracy and visual quality. Unlike 3DGS methods that produce blurry geometry and 2DGS variants that oversmooth textures, EGGS effectively combines both representations to recover sharp details and clean geometry, outperforming all baselines on Mip-NeRF360, LLFF, and Tanks & Temples.

EGGS delivers sharper and more accurate geometry than 3DGS while maintaining strong appearance quality. Compared to methods like 2DGS and SUGAR that focus solely on surface reconstruction, EGGS achieves a superior balance—producing clean, detailed depth maps and competitive geometry without sacrificing visual fidelity.

BibTeX

@inproceedings{
      anonymous2025eggs,
      title={{EGGS}: Exchangeable 2D/3D Gaussian Splatting for Geometry-Appearance Balanced Novel View Synthesis},
      author={Anonymous},
      booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
      year={2025},
      url={https://openreview.net/forum?id=25C8oC1pb2}
    }

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

Abstract

Videos

LLFF Dataset

Fern

Room

Trex

Tanks&Temples Dataset

Horse

Francis

Family

Barn

Mip-NeRF360 Dataset

Flowers

Room

Quantitative Comparisons

Comparison on appearance.

Comparison on geometry accuracy.

BibTeX