GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting

By Kai Zhang, Sai Bi, Hao Tan, Yuanbo Xiangli, Nanxuan Zhao, Kalyan Sunkavalli, Zexiang Xu. European Conference on Computer Vision, 2024

348 citations recorded by the source corpus.

We propose GS-LRM, a scalable large reconstruction model that can predict high-quality 3D Gaussian primitives from 2-4 posed sparse images in 0.23 seconds on single A100 GPU. Our model features a very simple transformer-based architecture; we patchify input posed images, pass the concatenated multi-view image tokens through a sequence of transformer blocks, and decode final per-pixel Gaussian parameters directly from these tokens for differentiable rendering. In contrast to previous LRMs that can only reconstruct objects, by predicting per-pixel Gaussians, GS-LRM naturally handles scenes with large variations in scale and complexity. We show that our model can work on both object and scene captures by training it on Objaverse and RealEstate10K respectively. In both scenarios, the models outperform state-of-the-art baselines by a wide margin. We also demonstrate applications of our model in downstream 3D generation tasks. Our project webpage is available at: https://sai-bi.github.io/project/gs-lrm/ .

Subfields: Computer Vision and Pattern Recognition, Industrial and Manufacturing Engineering

View Paper: https://www.semanticscholar.org/paper/8ed0477f640fa3a2d5411155e445d13752821629

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

VGGT: Visual Geometry Grounded Transformer (2025, neutral): “Feed-forward Novel View Synthesis is progressing rapidly [4, 31, 34, 36, 70, 84, 95, 104].”
CAT3D: Create Anything in 3D with Multi-View Diffusion Models (2024, background): “Another line of research is to learn feed-forward models that take a few views as input, and output 3D representations directly, without an optimization process per instance [6, 55, 56, 57, 58, 59, 60].”
MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (2024, background): “Very recently, a growing number of feed-forward 3DGS models [1, 2, 36, 37, 42, 51, 53] have been proposed to solve the sparse-view-to-3D task.”