Research map / 3D vision / NeRF and radiance fields
Do Flat Minima Improve Sparse Novel View Synthesis?
Youngsik Yun, Dongjun Gu, Youngjung Uh
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 |
|---|---|---|---|
| apply naive SAM to 3DGS (ablation) | 3DGS · LLFF (3 views) | PSNR: 19.810 -> 20.198 | Improved |
| add structure-aware regularization weight (ablation) | 3DGS + SAM · LLFF (3 views) | PSNR: 20.198 -> 20.570 | Improved |
| add structure-aware sharpness (ablation) | 3DGS + SAM · LLFF (3 views) | LPIPS: .2095 -> .2023 | Improved |
| add structure-aware regularization weight (ablation) | 3DGS + SAM + SAS · LLFF (3 views) | PSNR: 20.560 -> 20.783 | Improved |
| add structure-aware sharpness (ablation) | 3DGS + SAM + SAR · LLFF (3 views) | LPIPS: .2142 -> .1965 | Improved |
| add structure-aware sharpness regularization | NexusGS · MipNeRF-360 (12 views) | PSNR: 18.506 -> 18.736 | Improved |
| add method to Gaussian mean (ablation) | 3DGS + Ours · LLFF (3 views) | PSNR: 20.053 -> 20.783 | Improved |
| apply method only in late 12.5% of iterations (ablation) | 3DGS + Ours · LLFF (3 views) | training time: 2.77x -> 1.15x | Improved |
| apply method only in late 12.5% of iterations (ablation) | 3DGS + Ours · LLFF (3 views) | AVGE: 0.0132 -> 0.0095 | Hurt |
| use LoG instead of DoH tolerance map (ablation) | 3DGS + Ours · LLFF (3 views) | PSNR: 20.646 -> 20.783 | Improved |
| vary max regularization weight from 0.95 to 0.50 (ablation) | 3DGS + Ours · LLFF (3 views) | PSNR: 20.783 -> 20.730 | No clear effect |
| add method to RWP | 3DGS + RWP · MipNeRF-360 (12 views) | LPIPS: .3767 -> .3567 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- DropGaussian — DropGaussian: Structural Regularization for Sparse-view Gaussian Splatting (CVPR 2025) · better on 2 of 2 shared results
- NexusGS — NexusGS: Sparse View Synthesis with Epipolar Depth Priors in 3D Gaussian Splatting (CVPR 2025) · better on 2 of 2 shared results
- CoR-GS — CoR-GS: Sparse-View 3D Gaussian Splatting via Co-Regularization (ECCV 2024) · better on 2 of 2 shared results
- FreeNeRF — FreeNeRF: Improving Few-Shot Neural Rendering With Free Frequency Regularization (CVPR 2023) · better on 1 of 1 shared result
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
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.88)
- Seeking Consistent Flat Minima for Better Domain Generalization via Refining Loss Landscapes (CVPR 2025 · similarity 0.87)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.87)
- Improving Sharpness-Aware Minimization by Lookahead (ICML 2024 · similarity 0.87)
- GenRec: Knowing Where to Reconstruct and Where to Generate (NeurIPS 2026 · similarity 0.86)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.86)
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.
