文章预览
本期带来上周在arXiv公开的论文5篇,涉及3D重建、3D生成。 一、3D重建 1.[2024-ECCV] SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction 作者机构:Marko Mihajlovic, et al. ETH Zürich, etc. 论文地址: https://arxiv.org/pdf/2409.11211 项目地址: https://markomih.github.io/SplatFields/ 贡献: We propose a novel optimization strategy, named SplatFields, which introduces spatial bias into the 3D Gaussian Splatting technique to stabilize the
optimization process under sparse views. We extend our formulation to dynamic scenes, demonstrating superior reconstruction quality compared to recent state-of-the-art methods . We provide a detailed analysis of various modeling strategies, confirming the
optimality of our framework for the tasks of sparse multi-view reconstruction. 二、3D生成 1.[2024-arXiv] 3DTopia-XL: Scaling High-quality 3D Asset Generation via Primitive Diffusion 作者机构:Zhaoxi Chen, et al. Nanyang Technological U
………………………………