NeRF-Factory: An awesome PyTorch NeRF collection

Project Page | Checkpoints

Attention all NeRF researchers! We are here with a PyTorch-reimplemented large-scale NeRF library. Our library is easily extensible and usable.

This contains PyTorch-implementation of 7 popular NeRF models.

NeRF: [Project Page] [Paper] [Code]
NeRF++: [Paper] [Code]
DVGO: [Project Page] [Paper-v1] [Paper-v2] [Code]
Plenoxels: [Project Page] [Paper] [Code]
Mip-NeRF: [Project Page] [Paper] [Code]
Mip-NeRF360: [Project Page] [Paper] [Code]
Ref-NeRF: [Project Page] [Paper] [Code]

and also 7 popular NeRF datasets.

NeRF Blender: link
NeRF LLFF: link
Tanks and Temples: link
LF: link
NeRF-360: link
NeRF-360-v2: link
Shiny Blender: link

You only need to do for running the code is:

python3 -m run --ginc configs/[model]/[data].gin
# ex) python3 -m run --ginc configs/nerf/blender.gin

We also provide convenient visualizers for NeRF researchers.

Contributor

This project is created and maintained by Yoonwoo Jeong, Seungjoo Shin, and Kibaek Park.

Requirements

conda create -n nerf_factory -c anaconda python=3.8
conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch
pip3 install -r requirements.txt

## Optional(Plenoxel)
pip3 install .

## Or you could directly build from nerf_factory.yml
conda env create --file nerf_factory.yml

Command

python3 -m run --ginc configs/[model]/[data].gin
# ex) python3 -m run --ginc configs/nerf/blender.gin

Preparing Dataset

We provide an automatic download script for all datasets.

# NeRF-blender dataset
bash scripts/download_data.sh nerf_synthetic
# NeRF-LLFF(NeRF-Real) dataset
bash scripts/download_data.sh nerf_llff
# NeRF-360 dataset
bash scripts/download_data.sh nerf_real_360
# Tanks and Temples dataset
bash scripts/download_data.sh tanks_and_temples
# LF dataset
bash scripts/download_data.sh lf
# NeRF-360-v2 dataset
bash scripts/download_data.sh nerf_360_v2
# Shiny-blender dataset
bash scripts/download_data.sh shiny_blender

Run the Code!

A very simple script to run the code.

Training Code

A script for running the training code.

python3 run.py --ginc configs/[model]/[data].gin --scene [scene]

## ex) run training nerf on chair scene of blender dataset
python3 run.py --ginc configs/nerf/blender.gin --scene chair

Evaluation Code

A script for running the evaluation code only.

python3 run.py --ginc configs/[model]/[data].gin --scene [scene] \
--ginb run.run_train=False

## ex) run evaluating nerf on chair scene of blender dataset
python3 run.py --ginc configs/nerf/blender.gin --scene chair \
--ginb run.run_train=False

Custom

How to add the custom dataset and the custom model in NeRF-Factory?

Custom Dataset

Add files of the custom dataset on ./data/[custom_dataset].
Implement a dataset loader code on ./src/data/data_util/[custom_dataset].py.
Implement a custom dataset class LitData[custom_dataset] on ./src/data/litdata.py.
Add option of selecting the custom dataset on the function def select_dataset() of ./utils/select_option.py.
Add gin config file for each model as ./configs/[model]/[custom_dataset].gin.

Custom Model

Implement a custom model code on ./src/model/[custom_model]/model.py.
Implement a custom model's helper code on ./src/model/[custom_model]/helper.py.
[Optional] If you need more code files for the custom model, you can add them in ./src/model/[custom_model]/.- Add option of selecting the custom model on the function def select_model() of ./utils/select_option.py.
Add gin config file for each model as ./configs/[custom_model]/[dataset].gin.

kakaobrain / nerf-factory