Research map / 3D vision / NeRF and radiance fields
GAURA: Generalizable Approach for Unified Restoration and Rendering of Arbitrary Views
Vinayak Gupta, Rongali Simhachala Venkata Girish, Mukund Varma T, Ayush Tewari, Kaushik Mitra
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 degradation-aware latent modules (DLM) to feature UNet, view transformer, ray transformer (ablation) | GNT · Low-light (AlethNeRF) ablation | PSNR: 17.69 -> 19.37 | Improved |
| add DLM only to the feature UNet (ablation) | GNT · Low-light (AlethNeRF) ablation | PSNR: 17.69 -> 17.98 (SSIM 0.704 -> 0.680) | No clear effect |
| add adaptive residual module (ARM) to DLM (ablation) | GAURA · Low-light (AlethNeRF) ablation | PSNR: 19.37 -> 19.91 | Improved |
| use hypernetwork-style DLM with ARM instead of simple concatenation conditioning (ablation) | GAURA · Real-world motion blur | PSNR: 21.34 -> 22.12 | Improved |
| one generalizable degradation-aware model versus all-in-one restoration followed by GNT | GAURA · LLFF-Corrupted haze | PSNR: 15.81 -> 18.95 | Improved |
| generalizable GAURA versus per-scene 3D Restore on motion blur | GAURA · Real-world motion blur (DeblurNeRF, 10 scenes) | PSNR: 25.65 -> 22.12 | Hurt |
| generalizable GAURA versus per-scene 3D Restore on low-light | GAURA · Real-world low-light (AlethNeRF, 5 scenes) | PSNR: 17.64 -> 19.91 | Improved |
| fine-tune only a new latent code for unseen degradations | GAURA · Fine-tuned snow removal | PSNR: 21.96 -> 22.61 (vs Vanilla-GNT) | Improved |
| reduce input views from 10 to 3 (ablation) | GAURA · Low-light (AlethNeRF) | PSNR: 19.91 -> 19.18 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports GAURA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ReSplat: Degradation-agnostic Feed-forward Gaussian Splatting via Self-guided Residual Diffusion (ICLR 2026 · 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
- Boosting View Synthesis With Residual Transfer (CVPR 2022 · similarity 0.88)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.86)
- 3DGStream: On-the-Fly Training of 3D Gaussians for Efficient Streaming of Photo-Realistic Free-Viewpoint Videos (CVPR 2024 · similarity 0.85)
- RestorGS: Depth-aware Gaussian Splatting for Efficient 3D Scene Restoration (CVPR 2025 · similarity 0.85)
- GenFusion: Closing the Loop between Reconstruction and Generation via Videos (CVPR 2025 · similarity 0.85)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (ICLR 2024 · 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.
