Paper: RENI++: A Rotation-Equivariant, Scale-Invariant, Natural Illumination Prior
RENI++ is a nerfstudio extension. It requires CUDA 12.8, Python 3.12, PyTorch 2.x, tiny-cuda-nn, and nerfstudio. The recommended way to run it is via Docker or Apptainer (for HPC clusters).
Requires the NVIDIA Container Toolkit.
git clone https://github.com/JADGardner/ns_reni.git
cd ns_reniSet up data and model directories. Either create symlinks in the project root:
ln -s /path/to/datasets data
ln -s /path/to/pretrained-models model-storage
mkdir -p outputsOr set environment variables (in your shell or a .env file in the project root):
# .env
DATA_PATH=/path/to/datasets
MODEL_STORAGE_PATH=/path/to/pretrained-models
OUTPUTS_PATH=/path/to/outputsBuild and run:
# Build the image (compiles CUDA extensions — takes 20-40 min first time)
docker compose build research
# Start an interactive shell
docker compose run research bash
# Or train directly
docker compose run research ns-train reni --data /workspace/data/RENI_HDRInside the container, the project is mounted at /workspace with:
/workspace/data-- datasets/workspace/outputs-- training outputs/workspace/model-storage-- pretrained checkpoints
See the .apptainer/ directory for HPC/SLURM setup.
git clone https://github.com/JADGardner/ns_reni.git
cd ns_reni
# Configure host paths
cp .apptainer/.env.example .apptainer/.env
# Edit .apptainer/.env with your cluster's data/model/output paths
# Build the SIF image + overlay (submit as a SLURM job on HPC)
.apptainer/apptainer.sh build
# Register ns_reni code in the overlay
.apptainer/apptainer.sh install
# Interactive shell
.apptainer/apptainer.sh shell
# Run a command
.apptainer/apptainer.sh exec -- ns-train reni --help
# Verify the container
.apptainer/apptainer.sh exec -- python .apptainer/test_container.pygit clone https://github.com/JADGardner/ns_reni.git
cd ns_reni
conda create --name reni -y python=3.12
conda activate reni
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu128
conda install -c conda-forge colmap -y
sudo apt install libopenexr-dev # or: conda install -c conda-forge openexr
pip install ninja git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch
git clone --depth 1 https://github.com/nerfstudio-project/nerfstudio.git
cd nerfstudio && pip install -e . && cd ..
pip install -e .-lcuda not found- Solution:
ln -s {cuda directory}/lib/stubs/libcuda.so {cuda directory}/lib/libcuda.so
- Solution:
python3 scripts/download_data.py ./data/
python3 scripts/download_models.py ./output/model/This downloads the data and model where RENI++ expects them. The data and the model path can be changed in the config.