文章预览
本期带来上周在arXiv公开的论文10篇，涉及3D重建、3D生成。 一、3D重建 1.[2024-arXiv] LM-Gaussian: Boost Sparse-view 3D Gaussian Splatting with Large Model Priors 作者机构：Hanyang Yu, et al. The Hong Kong University of Science and Technology 论文地址： https://arxiv.org/pdf/2409.03456 项目地址： https://hanyangyu1021.github.io/lm-gaussian.github.io/ 贡献： We propose a new method capable of generating highquality novel views in a sparse-view setting with large model priors. Our method surpasses recent works in sparse-view settings, especially in large-scale 360-degree scenes. We introduce a Background-Aware Depth-guided Initialization Module, capable of simultaneously reconstructing high-quality dense point clouds and camera poses for initialization. We introduce a Multi-modal Regularized Gaussian Reconstruction Module that leverages regularization techniques from various domains to avoid overfitting issues. We present an Iterative Gaussian Refi
………………………………
