Follow this YouTube tutorial to understand the installation process more easily and if you have any questions feel free to join my discord and ask there.
Ever wanted to train a NeRF model of a fox in under 5 seconds? Or fly around a scene captured from photos of a factory robot? Of course you have!
Here you will find an implementation of four neural graphics primitives, being neural radiance fields (NeRF), signed distance functions (SDFs), neural images, and neural volumes. In each case, we train and render a MLP with multiresolution hash input encoding using the tiny-cuda-nn framework.
Instant Neural Graphics Primitives with a Multiresolution Hash Encoding
Thomas Müller, Alex Evans, Christoph Schied, Alexander Keller
arXiv:2201.05989 [cs.CV], Jan 2022
[ Project page ] [ Paper ] [ Video ] [ BibTeX ]
For business inquiries, please visit our website and submit the form: NVIDIA Research Licensing
- An NVIDIA GPU; tensor cores increase performance when available. All shown results come from an RTX 3090.
- Python ver: 3.9.*
- Visual Studio Community 2019 (Latest the best, ~8GB) Below are the install requirements
- CUDA v11.6. You can check ur CUDA version via
nvcc --version
in any prompt and if it's not CUDA11.6, refer to this to swap/install the correct version. - On some machines,
pyexr
refuses to install viapip
. This can be resolved by installing OpenEXR from here. See later. - This installation tutorial will be using Anaconda. Download anaconda prompt here.
- OptiX 7.3 or higher for faster mesh SDF training. You need to either login or join to obtain the installer. Set the system environment variables
OptiX_INSTALL_DIR
to the installation directory if it is not discovered automatically. Should look like this:
copy these files C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.6\extras\visual_studio_integration\MSBuildExtensions
to here C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\MSBuild\Microsoft\VC\v160\BuildCustomizations
cd
into a directory that you want to download the codes at. Eg. cd F:\Tutorial\ngp\
Begin by cloning this repository and all its submodules using the following command (if you don't have git, download here and add to path):
$ git clone --recursive https://github.com/nvlabs/instant-ngp
$ cd instant-ngp
if your python is not 3.9 (check with command python --version
) then you need to run the following command to get it to ver 3.9.*
conda install python=3.9
Then, open Developer Command Prompt, you can find this in your search bar.
Then cd
to where you cloned your repository so you are in its root folder /instant-ng/
:
cmake . -B build
cmake --build build --config RelWithDebInfo -j 16
If the any of these build fails, please consult this list of possible fixes before opening an issue.
If automatic GPU architecture detection fails, (as can happen if you have multiple GPUs installed), set the TCNN_CUDA_ARCHITECTURES
enivonment variable for the GPU you would like to use. The following table lists the values for common GPUs. If your GPU is not listed, consult this exhaustive list.
RTX 30X0 | A100 | RTX 20X0 | TITAN V / V100 | GTX 10X0 / TITAN Xp | GTX 9X0 | K80 |
---|---|---|---|---|---|---|
86 | 80 | 75 | 70 | 61 | 52 | 37 |
Install COLMAP, I used ver 3.7
Add it to your system environment variables at Environment Variables > System Variables Path > Edit environment variable
open anaconda prompt, if you don't have you don't have you can get it here
cd
into isntant-ngp as root
conda create -n ngp python=3.9
conda activate ngp
pip install -r requirements.txt
if pyexr
cannot be installed via pip install pyexr
, download OpenEXR‑1.3.2‑cp39‑cp39‑win_amd64.whl and move it to your root folder. Then you can run:
pip install OpenEXR-1.3.2-cp39-cp39-win_amd64.whl
Place your custom image set under data/<image_set_name>
Get transform.json
from the following command. Insert your path to your images at <image/path>
python scripts/colmap2nerf.py --colmap_matcher exhaustive --run_colmap --aabb_scale 16 --images <image/path>
transform.json
will be generated at the root folder, drag and drop it into your data/<image_set_name>
folder.
You have to reorganize the folder structure due to how transforms.json
is created...
For example:
File Structure BEFORE generating transform.json
📂instant-ngp/ # this is root
├── 📂data/
│ ├── 📂toy_truck/
│ │ ├── 📜toy_truck_001.jpg
│ │ ├── 📜toy_truck_002.jpg
│ │ │...
│ │...
│...
File Structure AFTER generating transform.json
📂instant-ngp/ # this is root
├── 📂data/
│ ├── 📂toy_truck/
│ │ ├── 📜transforms.json/
│ │ ├── 📂data/
│ │ │ ├── 📂toy_truck/
│ │ │ │ ├── 📜toy_truck_001.jpg
│ │ │ │ ├── 📜toy_truck_002.jpg
│ │ │ │ │...
│ │ │ │...
│ │ │...
│ │...
│...
Note: adjusting the "aabb_scale"
inside transform.json
can reduce load on GPU VRAM. The lower the value the less intensive it'll be.
Finally, to run instant-ngp:
<path_to_your_ngp>\instant-ngp\build\testbed.exe --scene data/<image_set_name>
eg.
C:\user\user\download\instant-ngp\build\testbed.exe --scene data/toy_truck
And it should launch the GUI and everything amazing with it, and my fork edits end here.
This codebase comes with an interactive testbed that includes many features beyond our academic publication:
- Additional training features, such as extrinsics and intrinsics optimization.
- Marching cubes for
NeRF->Mesh
andSDF->Mesh
conversion. - A spline-based camera path editor to create videos.
- Debug visualizations of the activations of every neuron input and output.
- And many more task-specific settings.
- See also our one minute demonstration video of the tool.
One test scene is provided in this repository, using a small number of frames from a casually captured phone video:
instant-ngp$ ./build/testbed --scene data/nerf/fox
Alternatively, download any NeRF-compatible scene (e.g. from the NeRF authors' drive). Now you can run:
instant-ngp$ ./build/testbed --scene data/nerf_synthetic/lego/transforms_train.json
For more information about preparing datasets for use with our NeRF implementation, please see this document.
instant-ngp$ ./build/testbed --scene data/sdf/armadillo.obj
instant-ngp$ ./build/testbed --scene data/image/albert.exr
To reproduce the gigapixel results, download, for example, the Tokyo image and convert it to .bin
using the scripts/image2bin.py
script. This custom format improves compatibility and loading speed when resolution is high. Now you can run:
instant-ngp$ ./build/testbed --scene data/image/tokyo.bin
Download the nanovdb volume for the Disney cloud, which is derived from here (CC BY-SA 3.0).
instant-ngp$ ./build/testbed --mode volume --scene data/volume/wdas_cloud_quarter.nvdb
To conduct controlled experiments in an automated fashion, all features from the interactive testbed (and more!) have Python bindings that can be easily instrumented.
For an example of how the ./build/testbed
application can be implemented and extended from within Python, see ./scripts/run.py
, which supports a superset of the command line arguments that ./build/testbed
does.
Happy hacking!
Before investigating further, make sure all submodules are up-to-date and try compiling again.
instant-ngp$ git submodule sync --recursive
instant-ngp$ git submodule update --init --recursive
If instant-ngp still fails to compile, update CUDA as well as your compiler to the latest versions you can install on your system. It is crucial that you update both, as newer CUDA versions are not always compatible with earlier compilers and vice versa. If your problem persists, consult the following table of known issues.
Problem | Resolution |
---|---|
CMake error: No CUDA toolset found / CUDA_ARCHITECTURES is empty for target "cmTC_0c70f" | Windows: the Visual Studio CUDA integration was not installed correctly. Follow these instructions to fix the problem without re-installing CUDA. (#18) |
Linux: Environment variables for your CUDA installation are probably incorrectly set. You may work around the issue using cmake . -B build -DCMAKE_CUDA_COMPILER=/usr/local/cuda-<your cuda version>/bin/nvcc (#28) |
|
CMake error: No known features for CXX compiler "MSVC" | Reinstall Visual Studio & make sure you run CMake from a developer shell. (#21) |
Compile error: undefined references to "cudaGraphExecUpdate" / identifier "cublasSetWorkspace" is undefined | Update your CUDA installation (which is likely 11.0) to 11.3 or higher. (#34 #41 #42) |
Compile error: too few arguments in function call | Update submodules with the above two git commands. (#37 #52) |
Python error: No module named 'pyngp' | It is likely that CMake did not detect your Python installation and therefore did not build pyngp . Check CMake logs to verify this. If pyngp was built in a different folder than instant-ngp/build , Python will be unable to detect it and you have to supply the full path to the import statement. (#43) |
If you cannot find your problem in the table, please feel free to open an issue and ask for help.
Many thanks to Jonathan Tremblay and Andrew Tao for testing early versions of this codebase and to Arman Toorians and Saurabh Jain for the factory robot dataset. We also thank Andrew Webb for noticing that one of the prime numbers in the spatial hash was not actually prime; this has been fixed since.
This project makes use of a number of awesome open source libraries, including:
- tiny-cuda-nn for fast CUDA MLP networks
- tinyexr for EXR format support
- tinyobjloader for OBJ format support
- stb_image for PNG and JPEG support
- Dear ImGui an excellent immediate mode GUI library
- Eigen a C++ template library for linear algebra
- pybind11 for seamless C++ / Python interop
- and others! See the
dependencies
folder.
Many thanks to the authors of these brilliant projects!
@article{mueller2022instant,
title = {Instant Neural Graphics Primitives with a Multiresolution Hash Encoding},
author = {Thomas M\"uller and Alex Evans and Christoph Schied and Alexander Keller},
journal = {arXiv:2201.05989},
year = {2022},
month = jan
}
Copyright © 2022, NVIDIA Corporation. All rights reserved.
This work is made available under the Nvidia Source Code License-NC. Click here to view a copy of this license.