nvblox Documentation#

nvblox is a library for real-time 3D reconstruction, designed for robotic applications.

With nvblox you can build, manipulate, and query reconstructions on the GPU. The library is highly optimized for runtime performance, taking full advantage of CUDA and NVIDIA hardware.

Depending on your use-case, you may interact with nvblox through either Python, C++ or ROS2.

Quickstart#

To install nvblox_torch via pip on a supported platform, run the following commands:

sudo apt-get install python3-pip libglib2.0-0 libgl1 # Open3D dependencies
pip3 install https://github.com/nvidia-isaac/nvblox/releases/download/v0.0.8/nvblox_torch-0.0.8rc5+cu12ubuntu24-863-py3-none-linux_x86_64.whl

See Installation for more options.

With nvblox installed, you’re ready to build your first reconstruction. Follow our 3D Reconstruction Example. After that try out some more functionality below.

Examples#

Below are some examples of building reconstructions using nvblox.

_images/3dmatch.gif _images/desk_radio_x2_600px.gif

3D Reconstruction

Deep Feature-based Reconstruction
_images/trajectory_optimization.gif _images/esdf_example.gif

Trajectory Optimization

Euclidean Signed Distance Field
_images/voxels_example.gif _images/nvblox_torch_realsense_live.gif

Voxels Access Example

Realsense Live Example

License#

This code is under an open-source license (Apache 2.0).

Papers#

If you find this library useful for your research, please consider citing our papers:

  • Alexander Millane, Helen Oleynikova, Emilie Wirbel, Remo Steiner, Vikram Ramasamy, David Tingdahl, and Roland Siegwart. “nvblox: GPU-Accelerated Incremental Signed Distance Field Mapping”. arXiv preprint arXiv:2311.00626 (2024).

  • Sundaralingam, Balakumar, Siva Kumar Sastry Hari, Adam Fishman, Caelan Garrett, Karl Van Wyk, Valts Blukis, Alexander Millane et al. “curobo: Parallelized collision-free minimum-jerk robot motion generation.”. arXiv preprint arXiv:2310.17274 (2023).

User's Guide

python

c++

Other