Research map / 3D vision / Gaussian splatting
CoSurfGS: 3D Surface Gaussian Splatting with Collaborative Distributed Learning for Large-scale Scene Reconstruction
Yuanyuan Gao, Yalun Dai, Hao Li, Weicai Ye, Junyi Chen, Danpeng Chen, Dingwen Zhang, Tong He, Guofeng Zhang, Junwei Han
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 Model Aggregation Scheme (distillation-based aggregation) (ablation) | CoSurfGS · Rubble | PSNR / SSIM / LPIPS: PSNR 18.07 -> 25.39; SSIM 0.441 -> 0.774; LPIPS 0.584 -> 0.267 | Improved |
| add Local Model Compression to local models (ablation) | CoSurfGS · Residence | memory / PSNR: 874MB -> 475MB; PSNR 22.13 -> 22.23 | Improved |
| increase prune percentage from 40% to 80% (ablation) | CoSurfGS · Rubble | memory / PSNR: 2.5GB -> 1.16GB; PSNR 25.19 -> 24.82 | Mixed |
| increase prune percentage from 40% to 60% (ablation) | CoSurfGS · Rubble | LPIPS: 0.285 -> 0.300 | Hurt |
| add geometry loss (no smooth weight) in device training (ablation) | device model · Rubble | PSNR: 25.03 -> 24.36 | Hurt |
| add smooth weight to geometry loss (ablation) | device model · Rubble | PSNR: 24.36 -> 24.73 | Improved |
| increase partition from (1*1)*4 to (3*3)*4 (ablation) | CoSurfGS · Rubble | PSNR / edge time: PSNR 25.37 -> 24.97; edge 130min -> 43min | Mixed |
| train unpartitioned (1*1)*1 whole scene (ablation) | CoSurfGS · Rubble | PSNR: OOM | Hurt |
| use distributed device-edge-cloud training vs prior surface methods | CoSurfGS · Mill19 and UrbanScene3D NVS | SSIM / PSNR / LPIPS: +0.1, +3.0, and +0.4 in SSIM, PSNR, and LPIPS | Improved |
| use CoSurfGS vs prior mesh methods | PGSR · BlendedMVS | F-score: +0.05 in the F-score metric average | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- NeUS — NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction (NeurIPS 2021) · better on 4 of 4 shared results
- DOGS — DOGS: Distributed-Oriented Gaussian Splatting for Large-Scale 3D Reconstruction Via Gaussian Consensus (NeurIPS 2024) · better on 0 of 2 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
- GigaGS: 3D Gaussian Based Planar Representation for Large-Scene Surface Reconstruction (AAAI 2025 · similarity 0.91)
- DOGS: Distributed-Oriented Gaussian Splatting for Large-Scale 3D Reconstruction Via Gaussian Consensus (NeurIPS 2024 · similarity 0.91)
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.91)
- Evolving High-Quality Rendering and Reconstruction in a Unified Framework with Contribution-Adaptive Regularization (CVPR 2025 · similarity 0.91)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.91)
- FlexGS: Train Once, Deploy Everywhere with Many-in-One Flexible 3D Gaussian Splatting (CVPR 2025 · 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.
