Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Plenoxels: Radiance Fields without Neural Networks
| Date | Stars |
|---|---|
| 2026-07-31 | 2890 |
| 2026-08-01 | 2890 |
| 2026-08-02 | 2890 |
| 2026-08-03 | 2890 |
| 2026-08-05 | 2890 |
| 2026-08-06 | 2890 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# Plenoxels: Radiance Fields without Neural Networks
Alex Yu\*, Sara Fridovich-Keil\*, Matthew Tancik, Qinhong Chen, Benjamin Recht, Angjoo Kanazawa
UC Berkeley
Website and video: <https://alexyu.net/plenoxels>
arXiv: <https://arxiv.org/abs/2112.05131>
[Featured at Two Minute Papers YouTube](https://youtu.be/yptwRRpPEBM) 2022-01-11
Despite the name, it's not strictly intended to be a successor of svox
Citation:
```
@inproceedings{yu2022plenoxels,
title={Plenoxels: Radiance Fields without Neural Networks},
author={Sara Fridovich-Keil and Alex Yu and Matthew Tancik and Qinhong Chen and Benjamin Recht and Angjoo Kanazawa},
year={2022},
booktitle={CVPR},
}
```
Note that the joint first-authors decided to swap the order of names between arXiv and CVPR proceedings.
This contains the official optimization code.
A JAX implementation is also available at <https://github.com/sarafridov/plenoxels>. However, note that the JAX version is currently feature-limited, running in about 1 hour per epoch and only supporting bounded scenes (at present).
### Examples use cases
Check out PeRFCeption [Jeong, Shin, Lee, et al], which uses Plenoxels with tuned parameters to generate a large
dataset of radiance fields:
https://github.com/POSTECH-CVLab/PeRFception
Artistic Radiance Fields by Kai Zhang et al
https://github.com/Kai-46/ARF-svox2
## Setup
**Windows is not officially supported, and we have only tested with Linux. Adding support would be welcome.**
First create the virtualenv; we recommend using conda:
```sh
conda env create -f environment.yml
conda activate plenoxel
```
Then clone the repo and install the library at the root (svox2), which includes a CUDA extension.
**If and only if** your CUDA toolkit is older than 11, you will need to install CUB as follows:
`conda install -c bottler nvidiacub`.
Since CUDA 11, CUB is shipped with the toolkit and installing this may lead to build errors.
To install the main library, simply run
```
pip install -e . --verbose
```
In the repo root directory.
## Getting datasets
We have backends for NeRF-Blender, LLFF, NSVF, and CO3D dataset formats, and the dataset will be auto-detected.
Please get the NeRF-synthetic and LLFF datasets from:
<https://drive.google.com/drive/folders/128yBriW1IG_3NJ5Rp7APSTZsJqdJdfc1>
(`nerf_synthetic.zip` and `nerf_llff_data.zip`).
We provide a processed Tanks and temples dataset (with background) in NSVF format at:
<https://drive.google.com/file/d/1PD4oTP4F8jTtpjd_AQjCsL4h8iYFCyvO/view?usp=sharing>
Note this data should be identical to that in NeRF++
Finally, the real Lego capture can be downloaded from:
https://drive.google.com/file/d/1PG-KllCv4vSRPO7n5lpBjyTjlUyT8Nag/view?usp=sharing
**Note: we currently do not support the instant-ngp format data (since the project was released before NGP). Using it will trigger the nerf-synthetic (Blender) data loader
due to similarity, but will not train properly. For real data we use the NSVF format.**
To convert instant-ngp data, please try our script
```
cd opt/scripts
python ingp2nsvf.py <ingp_data_dir> <output_data_dir>
```
## Optimization
For training a single scene, see `opt/opt.py`. The launch script makes this easier.
Inside `opt/`, run
`./launch.sh <exp_name> <GPU_id> <data_dir> -c <config>`
Where `<config>` should be `configs/syn.json` for NeRF-synthetic scenes,
`configs/llff.json`
for forward-facing scenes, and
`configs/tnt.json` for tanks and temples scenes, for example.
The dataset format will be auto-detected from `data_dir`.
Checkpoints will be in `ckpt/exp_name`.
**For pretrained checkpoints please see:** https://drive.google.com/drive/folders/1SOEJDw8mot7kf5viUK9XryOAmZGe_vvE?usp=sharing
## Evaluation
Use `opt/render_imgs.py`
Usage,
(in opt/)
`python render_imgs.py <CHECKPOINT.npExcerpt of 8,617 characters
Read on GitHub
163
3
Tom Byrer
1
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c51c80cb07a1460d, llm:Description: 'Plenoxels: Radiance Fields without Neural Networks' — implementation of neural radiance fields (NeRF) style view synthesis using non-neural (sparse voxel/plenoctree) representation. Language: Python. Topic: computer graphics / novel view synthesis.
matched fp:c51c80cb07a1460d, llm:Description: 'Plenoxels: Radiance Fields without Neural Networks' — implementation of neural radiance fields (NeRF) style view synthesis using non-neural (sparse voxel/plenoctree) representation. Language: Python. Topic: computer graphics / novel view synthesis.