Research map / 3D vision / Gaussian splatting
DiffusionHarmonizer: Bridging Neural Reconstruction and Photorealistic Simulation with Online Diffusion Enhancer
Yuxuan Zhang, Katarína Tóthová, Zian Wang, Kangxue Yin, Haithem Turki, Riccardo de Lutio, Yen-Yu Chang, Or Litany, Sanja Fidler, Zan Gojcic
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 temporal conditioning modules and temporal loss (ablation) | DiffusionHarmonizer · Novel trajectory simulation (in-domain) | temporal consistency: 0.9714 -> 0.9827 | Improved |
| add temporal conditioning modules and temporal loss (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | temporal consistency: 0.9502 -> 0.9670 | Improved |
| add temporal warping loss (ablation) | DiffusionHarmonizer · Novel trajectory simulation (in-domain) | temporal consistency: 0.9806 -> 0.9827 | Improved |
| add temporal warping loss (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | temporal consistency: 0.9618 -> 0.9670 | Improved |
| add artifact-correction data (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | FVD: 476.82 -> 453.17 | Improved |
| add ISP modification data (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | FVD: 465.7 -> 453.17 | Improved |
| add relighting data (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | FID: 102.15 -> 101.27 | Improved |
| add shadow simulation data (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | FID: 104.28 -> 101.27 | Improved |
| add asset re-insertion data (ablation) | DiffusionHarmonizer · Object insertion simulation (out-of-domain, Waymo) | FVD: 459.34 -> 453.17 | Improved |
| add multi-scale perceptual loss instead of no perceptual or LPIPS loss (ablation) | DiffusionHarmonizer · qualitative | visual quality | Improved |
| use single-step enhancer instead of Wan-Video V2V | DiffusionHarmonizer · Novel trajectory simulation (in-domain) | FVD: 506.86 -> 470.11 | Improved |
| use single-step enhancer instead of Wan-Video V2V | DiffusionHarmonizer · Novel trajectory simulation (in-domain) | inference time: 2827ms -> 212ms | Improved |
Closest papers by meaning
- Diffusion Renderer: Neural Inverse and Forward Rendering with Video Diffusion Models (CVPR 2025 · similarity 0.90)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.88)
- Diffusion Handles Enabling 3D Edits for Diffusion Models by Lifting Activations to 3D (CVPR 2024 · similarity 0.87)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.87)
- MirrorVerse: Pushing Diffusion Models to Realistically Reflect the World (CVPR 2025 · similarity 0.87)
- 3DGS-Enhancer: Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors (NeurIPS 2024 · similarity 0.87)
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.
