Research map / 3D vision / Gaussian splatting
EnerGS: Energy-Based Gaussian Splatting under Partial Geometric Priors
Rui Song, Tianhui Cai, Markus Gross, Yun Zhang, Walter Zimmer, Zhiyu Huang, Olaf Wysocki, Jiaqi Ma
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 |
|---|---|---|---|
| add occupied-space attraction energy (E_occ) (ablation) | EnerGS · KITTI | delta PSNR vs full model: -0.67 PSNR when removed | Improved |
| use decoupled geometry/appearance updates instead of joint optimization (ablation) | EnerGS · KITTI | delta PSNR vs full model: -0.56 PSNR when removed (KITTI); -3.20 (Waymo) | Improved |
| add free-space Boltzmann/softplus barrier (ablation) | EnerGS · Waymo Open Dataset | delta PSNR vs full model: -3.64 PSNR when removed | Improved |
| add weak prior in unknown space (E_unk) (ablation) | EnerGS · Waymo Open Dataset | delta PSNR vs full model: -3.62 PSNR when removed | Improved |
| use Welsch M-estimator instead of L2 for attraction (ablation) | EnerGS · KITTI | delta PSNR vs full model: -0.29 PSNR when replaced by L2 (KITTI); -3.60 (Waymo) | Improved |
| remove E_occ or decoupling (leakage and margin) (ablation) | EnerGS · KITTI | Leak: Leak -1.8 w/o E_occ, -2.8 w/o decoupling, but OccCov -4.5 and -4.0 | Mixed |
| make unknown-space attraction stronger than occupied attraction (R < 1) (ablation) | EnerGS · Waymo, OCC/UNK ratio sweep | delta PSNR: -0.66 at R=1/16; -0.29 at R=1/4 | Hurt |
| vary discrete pruning threshold tau_margin from 0.1 to 2.0 (ablation) | EnerGS · KITTI and Waymo | PSNR | No clear effect |
| scale sigma by 2x while fixing w/sigma^2 (ablation) | EnerGS · KITTI and Waymo | PSNR: OCC std/range 0.04 / 0.10 dB; UNK 0.05 / 0.12 dB | No clear effect |
| raise w_unk from 0.0625 to 1.0 (ablation) | EnerGS · KITTI and Waymo | PSNR | Hurt |
| add EnerGS regularization to SplatAD | SplatAD · KITTI | Leak: 22.1 -> 7.0 Leak; PSNR 17.14 -> 18.32 | Improved |
| add EnerGS regularization to SplatAD | SplatAD · Waymo | PSNR: 26.20 -> 26.34 PSNR; Leak 1.4 -> 1.2 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 4 on most shared results.
- Mip-Splatting — Mip-Splatting: Alias-free 3D Gaussian Splatting (CVPR 2024) · better on 2 of 4 shared results
- SplatAD — SplatAD: Real-Time Lidar and Camera Rendering with 3D Gaussian Splatting for Autonomous Driving (CVPR 2025) · better on 3 of 3 shared results
- Scaffold-GS — Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024) · better on 1 of 1 shared result
- GeoGaussian — GeoGaussian: Geometry-aware Gaussian Splatting for Scene Rendering (ECCV 2024) · better on 0 of 1 shared result
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
- Sparse2DGS: Geometry-Prioritized Gaussian Splatting for Surface Reconstruction from Sparse Views (CVPR 2025 · similarity 0.90)
- Metropolis-Hastings Sampling for 3D Gaussian Reconstruction (NeurIPS 2025 · similarity 0.89)
- HybridGS: Decoupling Transients and Statics with 2D and 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- DOGS: Distributed-Oriented Gaussian Splatting for Large-Scale 3D Reconstruction Via Gaussian Consensus (NeurIPS 2024 · similarity 0.89)
- IndoorGS: Geometric Cues Guided Gaussian Splatting for Indoor Scene Reconstruction (CVPR 2025 · similarity 0.89)
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.89)
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.
