NeRFStudio serves as an integrated platform for research and development in NeRF/3DGS domains. It provides a user-friendly interface with tools to streamline model construction, training, and evaluation processes. This guide documents the deployment experience and troubleshooting steps encountered during integration with a specific project.

Official Website: https://nerfstudio.orgGitHub Repository: https://github.com/nerfstudio-project/nerfstudoi

1. Ubuntu 22 Deployment and Error Handling

NeRFStudio deployment presents several challenges requiring NVIDIA GPU and CUDA support.

1.1 Conda Environment Setup

Python version must be 3.8 or higher. The following commands create an environment using Python 3.11:

conda create --name ns_env -y python=3.11
conda activate ns_env
pip install --upgrade pip

1.2 Dependency: Tiny-CUDA-NN

The framework depends on CUDA-enabled tiny-cuda-nn libray. For CUDA 11.8:

pip install torch==2.1.2+cu118 torchvision==0.16.2+cu118 --extra-index-url https://download.pytorch.org/whl/cu118
conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit
pip install ninja git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

For CUDA 12.x versions (current version 12.5):

export PATH="/usr/local/cuda/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda/lib64:$LD_LIBRARY_PATH"

Clone and build process:

git clone --recursive https://github.com/nvlabs/tiny-cuda-nn
cd tiny-cuda-nn
cmake . -B build -DCMAKE_BUILD_TYPE=RelWithDebInfo
cmake --build build --config RelWithDebInfo -j

Optional PyTorch extension installation:

pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

**Error Resolution 1:**When cloning repository, if encountering:

OSError: Unknown compute capability

Execute:

nvidia-smi --query-gpu=compute_cap --format=csv
export TCNN_CUDA_ARCHITECTURES=86  # Example for RTX 3090

**Error Resolution 2:**For:

CUDA_HOME environment variable is not set

Verify path configuration:

which nvcc
echo $LD_LIBRARY_PATH

**Error Resolution 3:**If encountering compiler errors during CMake:

conda install gxx

**Error Resolution 4:**For libcuda linking issues:

sudo ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/libcuda.so

1.3 NeRFStudio Installation

To install latest version:

git clone https://github.com/nerfstudio-project/nerfstudio.git
cd nerfstudio
pip install --upgrade pip setuptools
pip install -e .

2. SfM Implementation in NeRFStudio

3. 3DGS Reconstruction Techniques