sunset1995/DirectVoxGO
A neural radiance field method that replaces MLP-based NeRF with voxel grid representation for faster 3D scene reconstruction from images.
Not currently ranked — collecting fresh signals.
star history
DirectVoxGO optimizes a dense voxel grid directly to represent scene geometry and appearance, replacing the MLP backbone in NeRF for faster training and rendering. It uses dense 3D voxel grids for volume densities and 4D feature grids with shallow MLPs for view-dependent colors. The project includes custom CUDA extensions for additional speedup and implements an efficient O(N) distortion loss, with support for bounded and unbounded scene datasets.
Frequently asked
- What is sunset1995/DirectVoxGO?
- A neural radiance field method that replaces MLP-based NeRF with voxel grid representation for faster 3D scene reconstruction from images.
- Is DirectVoxGO open source?
- Yes — sunset1995/DirectVoxGO is an open-source project tracked on heatdrop.
- What language is DirectVoxGO written in?
- sunset1995/DirectVoxGO is primarily written in Python.
- How popular is DirectVoxGO?
- sunset1995/DirectVoxGO has 1.1k stars on GitHub.
- Where can I find DirectVoxGO?
- sunset1995/DirectVoxGO is on GitHub at https://github.com/sunset1995/DirectVoxGO.