nerf README

The project's README from github.com/bmild/nerf. As one markdown file: README.md.

NeRF: Neural Radiance Fields

Project Page | Video | Paper | Data

Open Tiny-NeRF in Colab
Tensorflow implementation of optimizing a neural representation for a single scene and rendering new views.

NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
Ben Mildenhall*1, Pratul P. Srinivasan*1, Matthew Tancik*1, Jonathan T. Barron2, Ravi Ramamoorthi3, Ren Ng1
1UC Berkeley, 2Google Research, 3UC San Diego
*denotes equal contribution
in ECCV 2020 (Oral Presentation, Best Paper Honorable Mention)

TL;DR quickstart

To setup a conda environment, download example training data, begin the training process, and launch Tensorboard:

conda env create -f environment.yml
conda activate nerf
bash download_example_data.sh
python run_nerf.py --config config_fern.txt
tensorboard --logdir=logs/summaries --port=6006

If everything works without errors, you can now go to localhost:6006 in your browser and watch the "Fern" scene train.

Setup

Python 3 dependencies:

The LLFF data loader requires ImageMagick.

We provide a conda environment setup file including all of the above dependencies. Create the conda environment nerf by running:

conda env create -f environment.yml

You will also need the LLFF code (and COLMAP) set up to compute poses if you want to run on your own real data.

What is a NeRF?

A neural radiance field is a simple fully connected network (weights are ~5MB) trained to reproduce input views of a single scene using a rendering loss. The network directly maps from spatial location and viewing direction (5D input) to color and opacity (4D output), acting as the "volume" so we can use volume rendering to differentiably render new views.

Optimizing a NeRF takes between a few hours and a day or two (depending on resolution) and only requires a single GPU. Rendering an image from an optimized NeRF takes somewhere between less than a second and ~30 seconds, again depending on resolution.

Running code

Here we show how to run our code on two example scenes. You can download the rest of the synthetic and real data used in the paper here.

Optimizing a NeRF

Run

bash download_example_data.sh

to get the our synthetic Lego dataset and the LLFF Fern dataset.

To optimize a low-res Fern NeRF:

python run_nerf.py --config config_fern.txt

After 200k iterations (about 15 hours), you should get a video like this at logs/fern_test/fern_test_spiral_200000_rgb.mp4:

ferngif

To optimize a low-res Lego NeRF:

python run_nerf.py --config config_lego.txt

After 200k iterations, you should get a video like this:

legogif

Rendering a NeRF

Run

bash download_example_weights.sh

to get a pretrained high-res NeRF for the Fern dataset. Now you can use render_demo.ipynb to render new views.

Replicating the paper results

The example config files run at lower resolutions than the quantitative/qualitative results in the paper and video. To replicate the results from the paper, start with the config files in paper_configs/. Our synthetic Blender data and LLFF scenes are hosted here and the DeepVoxels data is hosted by Vincent Sitzmann here.

Extracting geometry from a NeRF

Check out extract_mesh.ipynb for an example of running marching cubes to extract a triangle mesh from a trained NeRF network. You'll need the install the PyMCubes package for marching cubes plus the trimesh and pyrender packages if you want to render the mesh inside the notebook:

pip install trimesh pyrender PyMCubes

Generating poses for your own scenes

Don't have poses?

We recommend using the imgs2poses.py script from the LLFF code. Then you can pass the base scene directory into our code using --datadir <myscene> along with -dataset_type llff. You can take a look at the config_fern.txt config file for example settings to use for a forward facing scene. For a spherically captured 360 scene, we recomment adding the --no_ndc --spherify --lindisp flags.

Already have poses!

In run_nerf.py and all other code, we use the same pose coordinate system as in OpenGL: the local camera coordinate system of an image is defined in a way that the X axis points to the right, the Y axis upwards, and the Z axis backwards as seen from the image.

Poses are stored as 3x4 numpy arrays that represent camera-to-world transformation matrices. The other data you will need is simple pinhole camera intrinsics (hwf = [height, width, focal length]) and near/far scene bounds. Take a look at our data loading code to see more.

Citation

@inproceedings{mildenhall2020nerf,
  title={NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis},
  author={Ben Mildenhall and Pratul P. Srinivasan and Matthew Tancik and Jonathan T. Barron and Ravi Ramamoorthi and Ren Ng},
  year={2020},
  booktitle={ECCV},
}

nerf (bmild/nerf)

Searching Database…

11K ·GitHub
Nerf
What is this?The official TensorFlow implementation of NeRF, the ECCV 2020 system (best paper honorable mention) that learns a neural representation of a single scene and renders new views of it.
Is this a library I import, or research code I run myself?Research code: Python scripts (run_nerf.py) plus Jupyter notebooks for demos and mesh extraction. You clone the repo, build its conda environment, and launch training from the command line — there is no pip package.
How do I get it running?Create the nerf conda environment from the bundled setup file, run the provided script to download example training data, kick off training, and watch the Fern scene train in Tensorboard at localhost:6006.
What do I need installed before starting?Python 3 with TensorFlow 1.15, matplotlib, numpy, imageio and configargparse — the conda file covers those. The LLFF data loader additionally needs ImageMagick, and training on your own real scenes requires the LLFF code plus COLMAP for camera poses.
How long until I actually see a trained scene?Budget serious time: the example runs take 200k iterations, about 15 hours at low resolution. A NeRF is also tied to a single scene, so every new scene means another full training run.
Do I need a beefy GPU for this?Just one — the README says optimizing a NeRF only requires a single GPU. Rendering views from an optimized model is fast by comparison: under a second up to about 30 seconds per image.
What exactly is a neural radiance field?A small fully connected network (about 5MB of weights) that maps a 5D input — spatial location plus viewing direction — to color and opacity. Treated as a volume, it can be differentiably rendered to reproduce and synthesize new views of one scene.
Can I train it on my own photos?Yes. Compute camera poses with the imgs2poses.py script from the LLFF code (it uses COLMAP), then point the trainer at your folder with --datadir and --datasettype llff; for 360-degree captures add the --nondc --spherify --lindisp flags.
Can I get an actual 3D mesh out of a trained NeRF?Yes — extractmesh.ipynb runs marching cubes to pull a triangle mesh from the trained network. It needs the PyMCubes package, plus trimesh and pyrender if you want to render the mesh in the notebook.
Do I have to train from scratch to see results?No — the repo links a pretrained high-resolution NeRF for the Fern scene, and renderdemo.ipynb renders new views from it. For full paper-quality results, start from the configs in paperconfigs/.
What are similar projects?instant-ngp, nerfstudio, Open3D, TRELLIS.2, gaussian-splatting, DeepLearningExamples, Virgilio, vggt
Which topics is it in?3d-graphics, computer-graphics, numpy, tensorflow, rendering, jupyter notebook, 3d, computer-vision, deep-learning, research, gpu, machine-learning
Creator
bmild
Size
2.9 MB
Stars
11K
Updated
CommitApr 12
Age
6y
Commits · 90 Days