Research map / 3D vision / Gaussian splatting
ThermoGS: Decoupling Physical Surface Attributes for Spatio-Temporal Thermal Field Emulation via 4D Gaussian Splatting
Kun Yang, Yuxiang Liu, Zeyu Cui, Shen Yan, Maojun Zhang, Yu Liu, Xue Wang, Qing Wang
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 all five loss terms (edge, smooth, physical, ODE, PINN) instead of dropping any single one (ablation) | ThermoGS · Thermo S1-S4 | temperature MAE (C): 1.147 -> 0.932 | Improved |
| add all five loss terms instead of dropping any single one (ablation) | ThermoGS · Thermo S1-S4 | PSNR (dB): 31.17 -> 31.83 | Improved |
| predict physical attributes with PINN instead of table lookup (ablation) | ThermoGS · Thermo S1-S4 | temperature MAE (C): 1.282 -> 0.932 | Improved |
| predict physical attributes with PINN instead of table lookup (ablation) | ThermoGS · Thermo S1-S4 | PSNR (dB): 31.06 -> 31.83 | Improved |
| ThermoGS vs NTR-Gaussian on full-day cross-time/scene test | ThermoGS · Thermo S1-S4 | temperature MAE (C): 1.136 -> 0.932 | Improved |
| ThermoGS vs NTR-Gaussian on S3 held-out scenes | ThermoGS · Thermo S3 (S1,S2,S4) | temperature MAE (C): 3.122 -> 1.49 | Improved |
| ThermoGS vs NTR-Gaussian on S4 held-out scenes | ThermoGS · Thermo S4 (S1,S2,S3) | PSNR (dB): 28.80 -> 28.81 | No clear effect |
| ThermoGS vs 4DGS on S3 held-out scenes | ThermoGS · Thermo S3 (S1,S2,S4) | SSIM: 0.949 -> 0.955 | No clear effect |
| ThermoGS vs NTR-Gaussian on nighttime NTR dataset S2 | ThermoGS · NTR Dataset nighttime S2 | temperature MAE (C): 1.709 -> 0.785 | Improved |
| ThermoGS vs NTR-Gaussian on nighttime NTR dataset S4 | ThermoGS · NTR Dataset nighttime S4 | temperature MAE (C): 1.065 -> 0.896 | Improved |
| infer attributes with PINN rather than NTR-Gaussian approach: emissivity | ThermoGS · Thermo physical attribute inference | MAE of emissivity vs reference: 0.470 -> 0.018 | Improved |
| infer attributes: convective coefficient h | ThermoGS · Thermo physical attribute inference | MAE of h vs reference: 32.80 -> 0.534 | Improved |
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.
- NTR-Gaussian — NTR-Gaussian: Nighttime Dynamic Thermal Reconstruction with 4D Gaussian Splatting Based on Thermodynamics (CVPR 2025) · better on 12 of 12 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
- NTR-Gaussian: Nighttime Dynamic Thermal Reconstruction with 4D Gaussian Splatting Based on Thermodynamics (CVPR 2025 · similarity 0.89)
- ThermalGaussian: Thermal 3D Gaussian Splatting (ICLR 2025 · similarity 0.89)
- PhysIR-Splat: Physically Consistent Thermal Infrared Radiative Transfer in 3D Gaussian Splatting (CVPR 2026 · similarity 0.89)
- Thermo-JEPA: Learning a Geometry-Grounded Thermal World Model via Cross-Modal Privileged Masking (ECCV 2026 · similarity 0.88)
- DTGS: Physics-Embedded Dynamic Thermal 3D Reconstruction with Gaussian Splatting (NeurIPS 2026 · similarity 0.88)
- Neural Field Thermal Tomography: A Differentiable Physics Framework for Non-Destructive Evaluation (NeurIPS 2026 · similarity 0.84)
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.
