GeometrySticker: Enabling Ownership Claim of Recolorized Neural Radiance Fields (ECCV 2024)
Paper: arXiv
Project Page: https://kevinhuangxf.github.io/GeometrySticker/
git clone --branch main --single-branch https://github.com/kevinhuangxf/GeometrySticker.git
# create conda environment
conda create -n geosticker python=3.8
# install pytorch dependencies
pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 --index-url https://download.pytorch.org/whl/cu121
# install tiny-cuda
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch
# install torch-scatter
pip install torch-scatter -f https://data.pyg.org/whl/torch-2.1.0+cu121.html
# install other dependencies
pip install requirements.txt
Please download the Blender and LLFF datasets from this link: NeRF datasets.
Run GeometrySticker with NGP-based NeRF.
cd exp/ngp
Train a nerf model.
ROOT_DIR=path/to/Synthetic_NeRF
python train.py \
--root_dir $ROOT_DIR/Lego \
--exp_name Lego \
--num_epochs 30 --batch_size 16384 --lr 2e-2 --eval_lpips
Train GeometrySticker.
python train_geometrysticker.py \
--root_dir $ROOT_DIR/Lego_geosticker/ \
--exp_name Lego \
--lr 1e-4 \
--num_epochs 5 \
--weight_path ckpts/nsvf/Lego/epoch=29_slim.ckpt \
--downsample 0.25
# Evaluating on recoloring
ROOT_DIR=path/to/Synthetic_NeRF
python train_geometrysticker.py \
--root_dir $ROOT_DIR/Lego_geosticker \
--exp_name Lego_geosticker \
--weight_path ckpts/nsvf/Lego_geosticker/epoch=4_slim.ckpt \
--downsample 0.25 \
--val_only
@article{huang2024geometrysticker,
title = {GeometrySticker: Enabling Ownership Claim of Recolorized Neural Radiance Fields},
author = {Xiufeng Huang, Ka Chun Cheung, Simon See, Renjie Wan},
journal = {European Conference on Computer Vision (ECCV)},
year = {2024},
}