Research map / 3D vision / NeRF and radiance fields
Nerfbusters: Removing Ghostly Artifacts from Casually Captured NeRFs
Frederik Warburg, Ethan Weber, Matthew Tancik, Aleksander Holynski, Angjoo Kanazawa
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 visibility loss to Nerfacto (ablation) | Nerfacto · Nerfbusters dataset (12 scenes) | PSNR: 17.00 -> 17.81 | Improved |
| add DSDS local 3D diffusion prior on top of visibility loss (ablation) | Nerfacto + Vis · Nerfbusters dataset (12 scenes) | LPIPS: 0.3432 -> 0.2496 | Improved |
| add DSDS prior on top of visibility loss (ablation) | Nerfacto + Vis · Nerfbusters dataset (12 scenes) | mean abs. disparity difference: 1.041 -> 0.114 | Improved |
| add DSDS prior (ablation) | Nerfacto + Vis · Nerfbusters dataset (12 scenes) | coverage: 0.854 -> 0.630 | Hurt |
| add hand-crafted 3D sparsity loss on top of visibility loss (ablation) | Nerfacto + Vis · Nerfbusters dataset (12 scenes) | PSNR: 17.81 -> 17.81 | No clear effect |
| add 3D TV regularizer on top of visibility loss (ablation) | Nerfacto + Vis · Nerfbusters dataset (12 scenes) | mean normal error (degrees): 57.73 -> 61.93 | Hurt |
| add RegNeRF 2D TV regularizer on top of visibility loss (ablation) | Nerfacto + Vis · Nerfbusters dataset (12 scenes) | PSNR: 17.81 -> 17.49 | Hurt |
| use importance sampling by densities instead of uniform cube sampling (ablation) | DSDS · garbage scene | PSNR: 14.61 -> 16.46 | Improved |
| use densities instead of accumulation weights for cube importance sampling (ablation) | DSDS · garbage scene | PSNR: 15.86 -> 16.46 | Improved |
| use DSDS with thresholded densities instead of SDS with sigmoid or clamp activation (ablation) | diffusion prior loss · garbage scene | PSNR: 12.53 -> 15.86 | Improved |
| vary cube size range from 1-10% to 1-20% (ablation) | DSDS · garbage scene | PSNR: 15.86 -> 17.05 | Mixed |
Closest papers by meaning
- Optimize the Unseen - Fast NeRF Cleanup with Free Space Prior (NeurIPS 2025 · similarity 0.93)
- NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wild (CVPR 2024 · similarity 0.92)
- Aug-NeRF: Training Stronger Neural Radiance Fields With Triple-Level Physically-Grounded Augmentations (CVPR 2022 · similarity 0.91)
- Removing Objects From Neural Radiance Fields (CVPR 2023 · similarity 0.91)
- LuSh-NeRF: Lighting up and Sharpening NeRFs for Low-light Scenes (NeurIPS 2024 · similarity 0.90)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.90)
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.
