ACRF: Compressing Explicit Neural Radiance Fields via Attribute Compression (ICLR 2024)

This is the official implementation of ACRF, a NeRF compression algorithm based on 3D compression techniques. For technical details, please refer to:

ACRF: Compressing Explicit Neural Radiance Fields via Attribute Compression
Guangchi Fang, Qingyong Hu, Longguang Wang, Yulan Guo.
[Paper]

(1) Setup

This code has been tested with Python 3.7, torch 1.12.1, CUDA 11.6.

• Clone the repository

git clone git@github.com:fatPeter/ACRF.git && cd ACRF

• Setup python environment

conda create -n ACRF python=3.7
conda activate ACRF
pip install -r requirements.txt
pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 -f https://download.pytorch.org/whl/torch_stable.html
pip install torch-scatter==2.1.0 -f https://data.pyg.org/whl/torch-1.12.1%2Bcu116.html
cd lib/cuda
python setup.py install

• Download datasets: NeRF, T&T (masked).

  Directory structure (click to expand)

  data
  ├── nerf_synthetic     # Link: https://drive.google.com/drive/folders/128yBriW1IG_3NJ5Rp7APSTZsJqdJdfc1
  │   └── [chair|drums|ficus|hotdog|lego|materials|mic|ship]
  │       ├── [train|val|test]
  │       │   └── r_*.png
  │       └── transforms_[train|val|test].json
  │
  └── TanksAndTemple     # Link: https://dl.fbaipublicfiles.com/nsvf/dataset/TanksAndTemple.zip
      └── [Barn|Caterpillar|Family|Ignatius|Truck]
          ├── intrinsics.txt
          ├── rgb
          │   └── [0|1|2]_*.png
          └── pose
              └── [0|1|2]_*.txt
  

  For T&T, fix the intrinsics.txt of Ignatius following this issue.

(2) ACRF (full mode)

• Train, compress, decompress and eval scripts are in ./acrf:

cd acrf

• Train, compress, decompress and eval on Synthetic-NeRF:

python scripts.py --dataset syn --lamda 1e-2
# lamda: 1e-2, 5e-3, 2e-3

• Train, compress, decompress and eval on Tanks&Temples:

python scripts.py --dataset tnt --lamda 1e-2
# lamda: 1e-2, 5e-3, 2e-3

(3) ACRF-F (light mode)

• Train, compress, decompress and eval scripts are in ./acrf_f:

cd acrf_f

• Train, compress, decompress and eval on Synthetic-NeRF:

python scripts.py --dataset syn --Qstep 0.5
# Qstep: 0.5, 1, 2

• Train, compress, decompress and eval on Tanks&Temples:

python scripts.py --dataset tnt --Qstep 0.5
# Qstep: 0.5, 1, 2

Acknowledgement. This repository is originally based on VQRF and 3DAC.