Research map / 3D vision / Gaussian splatting
Gaussians on Fire: High-Frequency Reconstruction of Flames
Jakob Nazarenus, Dominik Michels, Wojciech Palubicki, Simin Kou, Fang-Lue Zhang, Soren Pirk, Reinhard Koch
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 flow-fused initialization of dynamic Gaussians (vs random) (ablation) | Ours · Real-world fire (17 scenes) | PSNR_flame: 27.55 -> 30.53 | Improved |
| add flow-fused initialization of dynamic Gaussians (vs random) (ablation) | Ours · Synthetic fire (velocity) | CosSim: 0.338 -> 0.744 | Improved |
| add hardware-synchronized timing (vs half-frame offset) (ablation) | Ours · Real-world fire (17 scenes) | PSNR_flame: 30.20 -> 30.53 | Improved |
| add hardware-synchronized timing (vs half-frame offset) (ablation) | Ours · Synthetic fire (velocity) | CosSim: 0.736 -> 0.744 | Improved |
| add alpha loss mask (ablation) | Ours · Real-world fire (17 scenes) | PSNR_flame: 30.78 -> 30.53 | Hurt |
| add alpha loss mask (ablation) | Ours · Synthetic fire (velocity) | CosSim: 0.697 -> 0.744 | Improved |
| add static pretraining (vs joint optimization) (ablation) | Ours · Real-world fire (17 scenes) | PSNR: 34.48 -> 37.54 | Improved |
| add static pretraining (vs joint optimization) (ablation) | Ours · Synthetic fire (velocity) | CosSim: 0.749 -> 0.744 | No clear effect |
| use exponential instead of linear transmittance (ablation) | Ours · Synthetic scenes (3, single time) | PSNR_flame: 18.51 -> 18.16 | 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 2 on most shared results.
- FreeTimeGS — FreeTimeGS: Free Gaussian Primitives at Anytime Anywhere for Dynamic Scene Reconstruction (CVPR 2025) · better on 6 of 6 shared results
- Grid4D — Grid4D: 4D Decomposed Hash Encoding for High-Fidelity Dynamic Gaussian Splatting (NeurIPS 2024) · better on 5 of 5 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
- HiCoM: Hierarchical Coherent Motion for Dynamic Streamable Scenes with 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- 3DFIRES: Few Image 3D REconstruction for Scenes with Hidden Surfaces (CVPR 2024 · similarity 0.87)
- DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos (NeurIPS 2025 · similarity 0.86)
- Gaussian-Flow: 4D Reconstruction with Dynamic 3D Gaussian Particle (CVPR 2024 · similarity 0.86)
- World from Motion: Generative Dynamic Gaussian Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.85)
- Learning Efficient Fuse-and-Refine for Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.85)
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.
