Download data from * data , unzip and put in the same folder with StyleRF (data and StyleRF are in the same folder) Download checkpoints for reconstructing density field from * log, unzip and put under StyleRF folder (log is under StyleRF)
pip install tqdm scikit-image opencv-python configargparse lpips imageio-ffmpeg kornia lpips tensorboard plyfile
cd StyleRF
There are 8 scences in nerf_synthetic data (chair, drums, hotdog, lego, materials, ship, ficus, mic), need to train a different model for each scene
To train model for a scene, run:
sh scripts/train_nerf.sh [scene] [GPU ID]
where scene is 1 of the 8 scene, GPU ID is the id of GPU to use
After training the model, to perform style transfer with custom style image, run:
sh scripts/test_style_nerf.sh [scene] [path to the style image] [GPU ID]
The generated images and video with style transfer are put under log_style/[scene]