Research map / 3D vision / Gaussian splatting
GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting
Kai Zhang, Sai Bi, Hao Tan, Yuanbo Xiangli, Nanxuan Zhao, Kalyan Sunkavalli, Zexiang Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| predict per-pixel 3D Gaussians with a plain transformer instead of triplane NeRF | LRM for objects · GSO (512 res) | PSNR: 26.54 -> 30.52 | Improved |
| use transformer predicting per-pixel Gaussians instead of LGM U-Net | object reconstruction · GSO (256 res) | PSNR: 21.44 -> 29.59 | Improved |
| use dense self-attention transformer instead of epipolar-based aggregation | scene reconstruction · RealEstate10K | PSNR: 25.89 -> 28.10 | Improved |
| use VGG-19 based perceptual loss instead of LPIPS (ablation) | GS-LRM training · training stability | training stability | Improved |
| clip Gaussian scale at maximum 0.3 (ablation) | GS-LRM training · training stability | training stability | Improved |
| pretrain at 256 then fine-tune at 512 resolution, and sample input and supervision views from same 8 renderings | GS-LRM training · convergence | convergence | Improved |
Later papers that compared with this method
24 later accepted papers report GS-LRM as a baseline in their results tables; 22 of them report a better number on most of the results they share with it.
- Free-Range Gaussians: Non-Grid-Aligned Generative 3D Gaussian Reconstruction (ECCV 2026 · better on 6 of 6 shared results)
- tttLRM: Test-Time Training for Long Context and Autoregressive 3D Reconstruction (CVPR 2026 · better on 4 of 4 shared results)
- UFO: Unifying Feed-Forward and Optimization-based Methods for Large Driving Scene Modeling (CVPR 2026 · better on 3 of 3 shared results)
- PhysGM: Large Physical Gaussian Model for Feed-Forward 4D Synthesis (CVPR 2026 · better on 2 of 3 shared results)
- TokenGS: Decoupling 3D Gaussian Prediction from Pixels with Learnable Tokens (CVPR 2026 · better on 2 of 3 shared results)
- SplatPainter: Interactive Authoring of 3D Gaussians from 2D Edits via Test-Time Training (ECCV 2026 · better on 3 of 3 shared results)
- PointSplat: Compact Gaussian Splatting via Human-Centric Prediction (ECCV 2026 · better on 3 of 3 shared results)
- DynamicVGGT: Learning Dynamic Point Maps for 4D Scene Reconstruction in Autonomous Driving (CVPR 2026 · better on 0 of 2 shared results)
16 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- pixelSplat — pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.92)
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025 · similarity 0.91)
- Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis (CVPR 2024 · similarity 0.90)
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.90)
- pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024 · similarity 0.90)
- DOGS: Distributed-Oriented Gaussian Splatting for Large-Scale 3D Reconstruction Via Gaussian Consensus (NeurIPS 2024 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
