文章预览
本期带来上周在arXiv公开的论文15篇，涉及3D重建、3D生成、3D编辑、4D重建。 一、3D重建 1.[2024-arXiv] HiSplat: Hierarchical 3D Gaussian Splatting for Generalizable Sparse-View Reconstruction 作者机构：Shengji Tang, et al. Fudan University, etc. 论文地址： https://arxiv.org/pdf/2410.06245 项目地址： https://open3dvlab.github.io/HiSplat/ 贡献： We first study and introduce hierarchical 3D Gassians representation in generalizable 3DGS. It can simultaneously reconstruct higher-quality large-scale structures and more delicate texture details to significantly alleviate mislocation and artefacts. We construct a novel generalizable framework named HiSplat, generating hierarchical 3D Gaussians to reconstruct the scene with only two-view reference images. Comprehensive experiments on various mainstream datasets demonstrate that, compared with previous methods, HiSplat obtains superior reconstruction quality and cross-dataset generalization. 2.[2024-a
………………………………
