Newest papers

KoalaGS: Complexity-Aware Continuous Level of Detail for Gaussian Splatting

Radu Beche, Raoul de Charette · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28615 · cited by 0 works (OpenAlex) · open-access PDF

Chinese

在收缩的计算预算下部署受过训练的3D Gaussian Splatting场景需要一个原始套件在更小的前端进行渲染时保持有效性。在这个连续的细节(LoD)设置中,挑战是决定哪个Gaussians应该保持容量,因为预算减少。对3DGS-MCMC重建的试点研究表明,从优化之前的输入视图计算的图像复杂性预测了重建错误集中在图像和3D空间中的位置。它还揭示了简单的地区含有太多的Gaussians,并在较低的预算中降低了复杂性,尽管其代表性更便宜。随后,在KoalaGS中,在培训过程中,我们使用这种复杂性优先将偏差迁移到不良重建的区域,同时确保在LoDs A

Machine translation (Meta M2M100); the English original follows.

Abstract

Deploying a trained 3D Gaussian Splatting scene under shrinking compute budgets requires one primitive set to remain effective when rendered at smaller prefixes. In this continuous level of detail (LoD) setting, the challenge is to decide which Gaussians should retain capacity as the budget decreases. A pilot study on 3DGS-MCMC reconstructions shows that image complexity, computed from the input views before optimization, predicts where reconstruction error concentrates in image and 3D space. It also reveals that simple regions contain too many Gaussians and degrade as much as complex ones at lower budgets, despite being inherently cheaper to represent. Subsequently, in KoalaGS, during training we use this complexity prior to bias relocation toward poorly reconstructed areas while ensuring smooth degradation across LoDs. A primitive importance score then enables inference at low-capacity, while preserving the scene structure. Tested across three datasets, KoalaGS maintains competitive full-capacity quality while significantly improving quality across the budget spectrum, yielding smoother transitions across LoDs, and enabling faster progressive streaming.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Artificial intelligence, Gaussian, Inference, Set (abstract data type), Computer vision

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange