You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
**Generatively edits NeRF scenes in a controlled and fast manner.**
16
+
17
+
SIGNeRF allows for generative 3D scene editing. We present a novel approach to combine [NeRFs](https://www.matthewtancik.com/nerf) as scene representation with the image diffusion model [StableDiffusion](https://github.com/Stability-AI/stablediffusion) to allow fast and controlled 3D generation.
SIGNeRF requires to use [Stable Diffusion Web UI](https://github.com/AUTOMATIC1111/stable-diffusion-webui). For detailed installation information follow our [installation guide](https://github.com/cgtuebingen/SIGNeRF?tab=readme-ov-file#installation).
28
+
29
+
## Running SIGNeRF
30
+
31
+
Details for running SIGNeRF can be found [here](https://github.com/cgtuebingen/SIGNeRF). Once installed, run:
|`signerf`| Full model, used in paper | 40min | Best |
42
+
|`signerf_nerfacto`| Faster model with Nerfacto | 20min | Good |
43
+
44
+
For more information on hardware requirements and training times, please refer to the [training section](https://github.com/cgtuebingen/SIGNeRF?tab=readme-ov-file#training-1).
SIGNeRF fully integrates into the Nerfstudio [viser](https://viser.studio) interface. It allows for easy editing of NeRF scenes with a few simple clicks. The user can select the editing method, the region to edit, and the object to insert. The reference sheet can be previewed and the NeRF is fine-tuned on the edited images. If you are interested in how we fully edit the Nerfstudio interface please find our code [here](https://github.com/cgtuebingen/SIGNeRF).
51
+
52
+
## Method
53
+
54
+
### Overview
55
+
56
+
SIGNeRF is a novel approach for fast and controllable NeRF scene editing and scene-integrated object generation. We introduce a new generative update strategy that ensures 3D consistency across the edited images, without requiring iterative optimization. We find that depth-conditioned diffusion models inherently possess the capability to generate 3D consistent views by requesting a grid of images instead of single views. Based on these insights, we introduce a multi-view reference sheet of modified images. Our method updates an image collection consistently based on the reference sheet and refines the original NeRF with the newly generated image set in one go. By exploiting the depth conditioning mechanism of the image diffusion model, we gain fine control over the spatial location of the edit and enforce shape guidance by a selected region or an external mesh.
57
+
58
+
For an in-depth visual explanation and our results please watch our [videos](https://www.youtube.com/playlist?list=PL5y23CB9WmildtW3QyMEi3arXg06zB4ex) or read our [paper](https://arxiv.org/abs/2401.01647).
We leverage the strengths of [ControlNet](https://github.com/lllyasviel/ControlNet), a depth condition image diffusion model, to edit an existing NeRF scene. We do so with a few simple steps in a single forward pass:
67
+
68
+
0. We start with an original NeRF scene and select an editing method / region
69
+
1. For object generation, we place a mesh object into the scene
70
+
2. And control the precise location and shape of the edit
71
+
3. We position reference cameras in the scene
72
+
4. Render the corresponding color, depth, and mask images, and arrange them into image grids
73
+
5. These grids are used to generate the reference sheet with conditioned image diffusion
74
+
6. Generate new edited images consistent with the reference sheet by leveraging an inpainting mask.
75
+
76
+
- Repeat step (6) for all cameras
77
+
78
+
7. Finally, the NeRF is fine-tuned on the edited images
SIGNeRF uses a novel technique called reference sheet generation. We observe that the image
85
+
diffusion model [ControlNet](https://github.com/lllyasviel/ControlNet) can already generate multiview consistent images of a scene without the need for iterative refinement like [Instruct-NeRF2NeRF](https://instruct-nerf2nerf.github.io/). While generating individual views sequentially introduces too much variation to integrate them into a consistent 3D model, arranging them in a grid of images that are processed by ControlNet in one pass significantly
86
+
improves the multi-view consistency. Based on the depth maps rendered from the original NeRF
87
+
scene we employ a depth-conditioned inpainting variant of ControlNet to generate such a reference sheet of the edited scene. A mask specifies the scene region where the generation should occur. This step gives a lot of control to the user. Different appearances can be produced by generating reference sheets with different seeds or prompts. The one sheet finally selected will directly determine the look of the final 3D scene.
88
+
89
+
If you want to learn more about the method, please read our [paper](https://arxiv.org/abs/2401.01647) or read the breakdown of the method in the [Radiance Fields article](https://radiancefields.com/controlnet-nerfs-signerf/).
90
+
91
+
## Related Work
92
+
93
+
SIGNeRF was inspired by the work of [Instruct-NeRF2NeRF](https://instruct-nerf2nerf.github.io/) which is also available in the [Nerfology](https://docs.nerf.studio/nerfology/methods/in2n.html) documentation.
94
+
95
+
## Authors
96
+
97
+
SIGNeRF was developed by [Jan-Niklas Dihlmann](https://jdihlmann.com) and [Andreas Engelhardt](https://aengelhardt.com/).
Copy file name to clipboardExpand all lines: docs/quickstart/data_conventions.md
+4
Original file line number
Diff line number
Diff line change
@@ -12,6 +12,10 @@ We use the OpenGL/Blender (and original NeRF) coordinate convention for cameras.
12
12
13
13
Our world space is oriented such that the up vector is +Z. The XY plane is parallel to the ground plane. In the viewer, you'll notice that red, green, and blue vectors correspond to X, Y, and Z respectively.
14
14
15
+
### Pixel coordinates
16
+
17
+
We assume coordinates correspond to the centers of pixels (e.g. generating a ray for pixel (0,0) will shoot the ray through the center of that pixel). This aligns more with graphics conventions, and is distinct from OpenCV where the corners are aligned with the pixel coordinate.
Copy file name to clipboardExpand all lines: nerfstudio/configs/external_methods.py
+17
Original file line number
Diff line number
Diff line change
@@ -14,6 +14,7 @@
14
14
15
15
16
16
"""This file contains the configuration for external methods which are not included in this repository."""
17
+
17
18
importinspect
18
19
importsubprocess
19
20
importsys
@@ -208,6 +209,22 @@ class ExternalMethod:
208
209
)
209
210
)
210
211
212
+
# SIGNeRF
213
+
external_methods.append(
214
+
ExternalMethod(
215
+
"""[bold yellow]SIGNeRF[/bold yellow]
216
+
For more information visit: https://docs.nerf.studio/nerfology/methods/signerf.html
217
+
218
+
To enable SIGNeRF, you must install it first by running:
219
+
[grey]pip install git+https://github.com/cgtuebingen/SIGNeRF[/grey] and install Stable Diffusion Web UI see [grey]https://github.com/cgtuebingen/SIGNeRF?tab=readme-ov-file#installation[/grey]""",
220
+
configurations=[
221
+
("signerf", "SIGNeRF method (high quality) used in paper"),
222
+
("signerf_nerfacto", "SIGNeRF method combined with Nerfacto (faster training less quality)"),
0 commit comments