Research map / 3D vision / Gaussian splatting
CylinderSplat: 3D Gaussian Splatting with Cylindrical Triplanes for Panoramic Novel View Synthesis
Qiwei Wang, Xianghui Ze, Jingyi Yu, Yujiao Shi
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 |
|---|---|---|---|
| use cylindrical triplane instead of Cartesian triplane in volume branch (ablation) | CylinderSplat volume branch · Matterport3D 2.0m two-view | WS-PSNR: 16.77 -> 22.17 | Improved |
| use cylindrical triplane instead of spherical triplane (ablation) | CylinderSplat volume branch · Matterport3D 2.0m two-view | WS-PSNR: 19.22 -> 22.17 | Improved |
| add volume branch to pixel branch (ablation) | CylinderSplat · Matterport3D 2.0m two-view | WS-PSNR: 23.21 -> 23.76 | Improved |
| add RGB retrieval for volume Gaussians (ablation) | cylindrical volume branch · Matterport3D 2.0m two-view | WS-PSNR: 20.16 -> 22.17 | Improved |
| use per-camera multiple triplanes instead of a single triplane (ablation) | CylinderSplat · Matterport3D 2.0m two-view | WS-PSNR: 23.47 -> 23.76 | Improved |
| use three-stage curriculum instead of end-to-end training (ablation) | CylinderSplat · Matterport3D 2.0m two-view | WS-PSNR: 23.25 -> 23.76 | Improved |
| increase triplane sampling density (16,32,64 -> 16,128,256) (ablation) | volume branch · Matterport3D 2.0m, volume only | WS-PSNR: 21.18 -> 23.88 (memory 15GB -> 25GB) | Improved |
| use isolated (per-camera) triplane initialization on real data instead of shared (ablation) | CylinderSplat · 360Loc | WS-PSNR: 27.33 -> 28.35 | Improved |
| use shared initialization on static synthetic data (ablation) | CylinderSplat · Matterport3D 2.0m | WS-PSNR: 23.24 -> 23.76 | Improved |
| increase input views from 2 to 4 (ablation) | CylinderSplat · 360Loc | WS-PSNR: 28.35 -> 28.97 | Improved |
| use CylinderSplat instead of MVSplat/PanSplat on near-baseline Residential and Replica | CylinderSplat · Residential 0.3m, Replica 1.0m | WS-PSNR: 28.25 vs 31.21 (Residential); 30.29 vs 30.78 (Replica) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- PanSplat — PanSplat: 4K Panorama Synthesis with Feed-Forward Gaussian Splatting (CVPR 2025) · better on 5 of 7 shared results
- Splatter360 — Splatter-360: Generalizable 360 Gaussian Splatting for Wide-baseline Panoramic Images (CVPR 2025) · better on 4 of 4 shared results
- OmniScene — Omni-Scene: Omni-Gaussian Representation for Ego-Centric Sparse-View Scene Reconstruction (CVPR 2025) · better on 3 of 3 shared results
- MVSplat — MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (ECCV 2024) · better on 1 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
- Splatter-360: Generalizable 360 Gaussian Splatting for Wide-baseline Panoramic Images (CVPR 2025 · similarity 0.93)
- PanSplat: 4K Panorama Synthesis with Feed-Forward Gaussian Splatting (CVPR 2025 · similarity 0.90)
- 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes (CVPR 2025 · similarity 0.90)
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 2025 · similarity 0.89)
- HiSplat: Hierarchical 3D Gaussian Splatting for Generalizable Sparse-View Reconstruction (ICLR 2025 · similarity 0.89)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · 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.
