用网格面约束高斯点，单目动态物体重建更稳定
Tessellation GS: Neural Mesh Gaussians for Robust Monocular Reconstruction of Dynamic Objects
- 以网格面为基底，分区域约束2D高斯点分布
- 在动态场景中降低29.1%的外观误差，49.2%的几何误差
- 适合单相机拍摄下的复杂动态物体重建
3D高斯泼溅（GS）虽能从带位姿图像序列生成高度逼真的三维场景，但因各向异性导致视角外推能力差，易过拟合且泛化性弱，尤其在稀疏视角和动态场景下表现不佳。本文提出基于网格面结构的分块高斯泼溅（Tessellation GS），实现单个连续移动或静止相机对动态场景的重建。方法将2D高斯点限制在局部区域，并通过网格面上的分层神经特征推断其属性；高斯细分由细节感知损失函数驱动的自适应面细分策略指导。此外，利用重建基础模型的先验初始化高斯形变，使优化方法在单静态相机下也能稳健重建通用动态物体。实验表明，本方法优于当前最优方法，在外观重建上降低29.1% LPIPS，几何重建上降低49.2% Chamfer距离。
原文摘要 · Abstract (English)
3D Gaussian Splatting (GS) enables highly photorealistic scene reconstruction from posed image sequences but struggles with viewpoint extrapolation due to its anisotropic nature, leading to overfitting and poor generalization, particularly in sparse-view and dynamic scene reconstruction. We propose Tessellation GS, a structured 2D GS approach anchored on mesh faces, to reconstruct dynamic scenes from a single continuously moving or static camera. Our method constrains 2D Gaussians to localized regions and infers their attributes via hierarchical neural features on mesh faces. Gaussian subdivision is guided by an adaptive face subdivision strategy driven by a detail-aware loss function. Additionally, we leverage priors from a reconstruction foundation model to initialize Gaussian deformations, enabling robust reconstruction of general dynamic objects from a single static camera, previously extremely challenging for optimization-based methods. Our method outperforms previous SOTA method, reducing LPIPS by 29.1% and Chamfer distance by 49.2% on appearance and mesh reconstruction tasks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。