VoxGRAF
This repository contains official code for the paper VoxGRAF: Fast 3D-Aware Image Synthesis with Sparse Voxel Grids.
You can find detailed usage instructions for training your own models and using pre-trained models below.
If you find our code or paper useful, please consider citing
@inproceedings{Schwarz2022NEURIPS,
title = {VoxGRAF: Fast 3D-Aware Image Synthesis with Sparse Voxel Grids},
author = {Schwarz, Katja and Sauer, Axel and Niemeyer, Michael and Liao, Yiyi and Geiger, Andreas},
booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
year = {2022}
}
Installation
First you have to make sure that you have all dependencies in place. The simplest way to do so, is to use anaconda.
You can create and activate an anaconda environment called voxgraf using
conda env create -f environment.yml
conda activate voxgraf
CUDA extension installation
Install pre-compiled CUDA extensions by running
./scripts/build_wheels.sh
Or install them individually by running
pip install dist/stylegan3_cuda-0.0.0-cp39-cp39-linux_x86_64.whl
pip install dist/svox2-voxgraf-0.0.1.dev0+sphtexcub.lincolor.fast-cp39-cp39-linux_x86_64.whl
pip install dist/MinkowskiEngine-0.5.4-cp39-cp39-linux_x86_64.whl # optional, only required when training with minkowski sparse convolutions
In case the wheels do not work for you, you can also install the extensions from source. For this please check the original repos: Stylegan-3, Minkowski Engine and for our version of Plenoxels follow the instructions here.
Pretrained models
To download the pretrained models run
./scripts/download_pretrained_models.sh
Evaluate pretrained models
# generate a video with 1x2 samples and interpolations between 2 keyframes each
python gen_video.py --network pretrained_models/ffhq256.pkl --seeds 0-3 --grid 1x2 --num-keyframes 2 --output ffhq_256_samples/video.mp4 --trunc=0.5
# generate grids of 3x4 samples and their depths
python gen_images.py --network pretrained_models/ffhq256.pkl --seeds 0-23 --grid 3x4 --outdir ffhq_256_samples --save_depth true --trunc=0.5
Train custom models
Download the data
Download FFHQ, AFHQ and Carla.
Preparing the data
To prepare the data at the required resolutions you can run
./scripts/make_dataset.sh /PATH/TO/IMAGES data/{DATASET_NAME}.json data/{DATASET_NAME} 32,64,128,256
This will create the datasets in data/{DATASET_NAME}_{RES}.zip.
Train models progressively
# Train a model on FFHQ progressively starting at image resolution 32x32 with voxel grid resolution 32x32x32
python train.py --outdir training-runs --gpus 8 --data data/ffhq_32.zip --batch 64 --grid-res 32
python train.py --outdir training-runs --gpus 8 --data data/ffhq_64.zip --batch 64 --grid-res 32 --resume /PATH/TO/32-IMG-32-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_64.zip --batch 64 --grid-res 64 --resume /PATH/TO/64-IMG-32-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_128.zip --batch 64 --grid-res 64 --lambda_vardepth 1e-3 --resume /PATH/TO/64-IMG-64-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_128.zip --batch 32 --grid-res 128 --lambda_vardepth 1e-3 --resume /PATH/TO/128-IMG-64-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_256.zip --batch 32 --grid-res 128 --lambda_vardepth 1e-3 --resume /PATH/TO/128-IMG-128-GRID-MODEL # Next stage
# Train a model on Carla at image resolution 32x32 with voxel grid resolution 32x32x32
python train.py --outdir training-runs --gpus 8 --data data/ffhq_32.zip --batch 64 --grid-res 32 --n-refinement 0 --use_bg False --lambda_sparsity 1e-8
python train.py --outdir training-runs --gpus 8 --data data/ffhq_64.zip --batch 64 --grid-res 32 --n-refinement 0 --use_bg False --lambda_sparsity 1e-8 --resume /PATH/TO/32-IMG-32-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_64.zip --batch 64 --grid-res 64 --n-refinement 0 --use_bg False --lambda_sparsity 1e-8 --resume /PATH/TO/64-IMG-32-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_128.zip --batch 64 --grid-res 64 --n-refinement 0 --use_bg False --lambda_sparsity 1e-8 --lambda_vardepth 1e-3 --resume /PATH/TO/64-IMG-64-GRID-MODEL # Next stage
python train.py --outdir training-runs --gpus 8 --data data/ffhq_128.zip --batch 32 --grid-res 128 --n-refinement 0 --use_bg False --lambda_sparsity 1e-8 --lambda_vardepth 1e-3 --resume /PATH/TO/128-IMG-64-GRID-MODEL # Next stage