Research map / 3D vision / NeRF and radiance fields
ViCA-NeRF: View-Consistency-Aware 3D Editing of Neural Radiance Fields
Jiahua Dong, Yu-Xiong 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 |
|---|---|---|---|
| use ViCA-NeRF (depth-guided projection plus blending) instead of Instruct-NeRF2NeRF | NeRF text editing · Difficult edits | text-image direction score: 0.1016 -> 0.1412 | Improved |
| use ViCA-NeRF instead of Instruct-NeRF2NeRF | NeRF text editing · Difficult edits | consistency score: 0.5161 -> 0.6498 | Improved |
| use ViCA-NeRF instead of Instruct-NeRF2NeRF | NeRF text editing · Simple edits | consistency score: 0.9004 -> 0.9035 | Improved |
| use ViCA-NeRF instead of Instruct-NeRF2NeRF | NeRF text editing · Simple edits | text-image direction score: 0.1477 -> 0.1545 | Improved |
| avoid iterative dataset updates by editing key views and projecting (editing time) | Face scene, 10,000 iterations on RTX 4090 · Face scene | editing time: 45 minutes -> 15 minutes | Improved |
| add blending refinement with averaged latent codes (ablation) | ViCA-NeRF · qualitative (face/wall scenes) | edit quality and consistency | Improved |
| add post-refinement after NeRF training (ablation) | ViCA-NeRF · qualitative | consistency | No clear effect |
| add warm-up iterations (0 to 10 or 30) (ablation) | ViCA-NeRF · qualitative | editing scale | Improved |
| set key view selection phi to 1 (almost all views as key views) (ablation) | ViCA-NeRF · qualitative | efficiency | Hurt |
Later papers that compared with this method
5 later accepted papers report ViCA-NeRF as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Mastering Regional 3DGS: Locating, Initializing, and Editing with Diverse 2D Priors (IJCV 2026 · better on 3 of 4 shared results)
- Fast Multi-view Consistent 3D Editing with Video Priors (AAAI 2026 · better on 2 of 2 shared results)
- GaussCtrl: Multi-View Consistent Text-Driven 3D Gaussian Splatting Editing (ECCV 2024 · better on 2 of 3 shared results)
- DATENeRF: Depth-Aware Text-based Editing of NeRFs (ECCV 2024 · better on 2 of 2 shared results)
- DGE: Direct Gaussian 3D Editing by Consistent Multi-view Editing (ECCV 2024 · better on 2 of 2 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
- Editing Conditional Radiance Fields (ICCV 2021 · similarity 0.91)
- Blending-NeRF: Text-Driven Localized Editing in Neural Radiance Fields (ICCV 2023 · similarity 0.91)
- SINE: Semantic-Driven Image-Based NeRF Editing With Prior-Guided Editing Field (CVPR 2023 · similarity 0.90)
- LAENeRF: Local Appearance Editing for Neural Radiance Fields (CVPR 2024 · similarity 0.89)
- Language-driven Object Fusion into Neural Radiance Fields with Pose-Conditioned Dataset Updates (CVPR 2024 · similarity 0.89)
- InNeRF360: Text-Guided 3D-Consistent Object Inpainting on 360-degree Neural Radiance Fields (CVPR 2024 · similarity 0.89)
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.
