LiTe-GS：面向3D高斯泼溅的Oracle高效下一最佳视角选择
原标题：LiTe-GS: Oracle-Efficient Next Best View Selection for 3D Gaussian Splatting
AI 摘要
论文提出 LiTe-GS，一种面向 3D Gaussian Splatting 的 oracle 高效下一最佳视角选择方法。它通过对候选视角进行随机子集评估，而非对全部候选池穷举打分，从而将 Fisher oracle 评估次数的期望复杂度降为与候选视角数量相关、与选择基数无关，并在理论上给出 oracle 复杂度与近似质量的保证。在 Blender 和 Mip-NeRF 360 上的实验表明，LiTe-GS 在保持与基于 Fisher 信息的基线相当的重建质量的同时，显著减少了 Fisher oracle 评估次数。
正文节选
LiTe-GS: Oracle-Efficient Next Best View Selection for 3D Gaussian Splatting Abstract Selecting informative camera views is critical for efficient training and adaptive refinement in 3D Gaussian Splatting, where each observation significantly influences model parameters. However, information-driven view-selection strategies can require repeated evaluations of expensive information-gain oracles as the number of candidate views increases. We propose LiTe-GS, an oracle-efficient method for next bes