Research map / 3D vision / 3D generation
DATENeRF: Depth-Aware Text-based Editing of NeRFs
Sara Rojas Martinez, Julien Philip, Kai Zhang, Sai Bi, Fujun Luan, Bernard Ghanem, Kalyan Sunkavalli
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 projection inpainting to depth-conditioned ControlNet editing (ablation) | DATENeRF · NeRF text editing (24 edits) | CLIP Direction Consistency: 0.6879 -> 0.7069 | Improved |
| add projection inpainting to depth-conditioned ControlNet editing (ablation) | DATENeRF · NeRF text editing (24 edits) | CLIP Text-Image Direction Similarity: 0.1772 -> 0.1866 | Improved |
| replace Instruct-Pix2Pix with ControlNet in Instruct-NeRF2NeRF | Instruct-NeRF2NeRF · NeRF text editing (24 edits) | CLIP Text-Image Direction Similarity: 0.1407 -> 0.1330 (similarity), 0.6349 -> 0.6799 (consistency) | Mixed |
| use projection inpainting with Instruct-Pix2Pix instead of plain Instruct-NeRF2NeRF (ablation) | Instruct-NeRF2NeRF · NeRF text editing (24 edits) | CLIP Direction Consistency: 0.6349 -> 0.6910 | Improved |
| use ControlNet without projection inpainting versus with it (ablation) | DATENeRF · NeRF text editing (24 edits) | CLIP Direction Consistency: 0.7069 -> 0.6879 | Hurt |
| keep reprojected pixels for all 20 denoising steps (N=20) instead of 5 (ablation) | projection inpainting · bear scene (qualitative) | visual quality | Hurt |
| no projection (N=0), depth-conditioned inpainting only (ablation) | projection inpainting · bear scene (qualitative) | photometric consistency | Hurt |
| edit all images upfront with projection inpainting instead of Iterative Dataset Update | DATENeRF · bear scene convergence | image edits and NeRF iterations to converge: 87 edits and 400 iterations vs 300 edits and 3000 iterations (IN2N) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ViCA-NeRF — ViCA-NeRF: View-Consistency-Aware 3D Editing of Neural Radiance Fields (NeurIPS 2023) · 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
- ED-NeRF: Efficient Text-Guided Editing of 3D Scene With Latent Space NeRF (ICLR 2024 · similarity 0.92)
- NeRFDeformer: NeRF Transformation from a Single View via 3D Scene Flows (CVPR 2024 · similarity 0.88)
- Customize your NeRF: Adaptive Source Driven 3D Scene Editing via Local-Global Iterative Training (CVPR 2024 · similarity 0.88)
- Perturb-and-Revise: Flexible 3D Editing with Generative Trajectories (CVPR 2025 · similarity 0.88)
- GenN2N: Generative NeRF2NeRF Translation (CVPR 2024 · similarity 0.88)
- DynVideo-E: Harnessing Dynamic NeRF for Large-Scale Motion- and View-Change Human-Centric Video Editing (CVPR 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.
