• Project Page
• T-RO 25 Paper, ICRA 24 Paper
• T-RO 25 Video, ICRA 24 Video
• Dataset

(Dec 2025) SiLVR 2.0 Code release for the T-RO version with epistemic uncertainty estimation.

(Sep 2025) SiLVR with uncertainty estimation has been accepted by the IEEE Transactions on Robotics (T-RO) and it will appear in the Special Section on Visual SLAM. New code will be released also here.

Build a docker image to install dependencies and run SiLVR.

mkdir data outputs
docker compose -f .docker/docker_compose.yaml run --build silvr

Tip 1: Update UID and GID in the .docker/.env to be the same as your system user ID (check by running id in the terminal) such that the host machine has proper permission to the data and outputs folders generated inside the docker.

Tip 2: You can use your own CUDA_ARCHITECTURES in the Dockerfile to make the tinycudann build quicker.

You can also install SiLVR to your system manually.

# install Pytorch
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118

# install Tiny-CUDA-NN
pip install ninja git+https://github.com/NVlabs/tiny-cuda-nn.git@v1.6#subdirectory=bindings/torch

# clone the repo
git clone https://github.com/ori-drs/silvr.git
cd silvr
pip install -e .

python scripts/data_downloader.py

python scripts/main.py --config configs/2024-03-13-roq-01-unc.yaml

python scripts/main.py --config configs/2024-bodleian-01+02-unc.yaml

Download sample data from Hugging face, setup the config file, and then run the training script.

python scripts/data_downloader.py
python scripts/train.py --config config/2024-03-13-roq-01.yaml

To reproduce the demo video which involves longer training on a bigger dataset, run the following training config.

python scripts/train.py --config config/2023-09-02-roq-hbac.yaml

We use Ruff as the formatter and linter. Install pre-commit will fix your code when you commit.

pre-commit install

If you found this software package useful, please consider citing our paper as

@article{tao2025silvr,
  title={SiLVR: Scalable Lidar-Visual Radiance Field Reconstruction with Uncertainty Quantification},
  author={Tao, Yifu and Fallon, Maurice},
  journal={IEEE Transactions on Robotics}, 
  year={2025},
}