Training Backend Setup

This guide covers the integrated training panel inside the GS Capture addon.

Data Requirements

Your training data folder must contain:

• images/ with image_0000.* files
• One of the following:
• transforms.json (Nerfstudio and gsplat)
• sparse/0/ COLMAP files (3DGS and GS-Lightning)

3D Gaussian Splatting (Original)

1. Clone the repository

   git clone https://github.com/graphdeco-inria/gaussian-splatting.git
   cd gaussian-splatting
   

2. Create and activate a conda environment

   conda create -n gaussian_splatting python=3.8
   conda activate gaussian_splatting
   

3. Install dependencies

   pip install -r requirements.txt
   pip install submodules/diff-gaussian-rasterization
   pip install submodules/simple-knn
   

4. In Blender preferences, set:
5. 3D Gaussian Splatting Path: the repo folder
6. 3D Gaussian Splatting Env: gaussian_splatting (or your env name)

Nerfstudio (splatfacto)

1. Create and activate a conda environment

   conda create -n nerfstudio python=3.10
   conda activate nerfstudio
   

2. Install PyTorch (adjust CUDA version as needed)

   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118
   

3. Install Nerfstudio

   pip install nerfstudio
   

4. In Blender preferences, set:
5. Nerfstudio Env: nerfstudio (or your env name)

GS-Lightning

1. Clone the repository

   git clone https://github.com/yzslab/gaussian-splatting-lightning.git
   cd gaussian-splatting-lightning
   

2. Create and activate a conda environment

   conda create -n gs_lightning python=3.10
   conda activate gs_lightning
   

3. Install PyTorch (adjust CUDA version as needed)

   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118
   

4. Install requirements

   pip install -r requirements.txt
   

5. In Blender preferences, set:
6. GS-Lightning Path: the repo folder
7. GS-Lightning Env: gs_lightning (or your env name)
8. In GS Capture settings, enable Export Object Masks and set Mask Format to GS-Lightning.

gsplat

1. Create and activate a conda environment

   conda create -n gsplat python=3.10
   conda activate gsplat
   

2. Install PyTorch (adjust CUDA version as needed)

   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118
   

3. Install gsplat

   pip install gsplat
   

4. Clone gsplat and locate the examples directory

   git clone https://github.com/nerfstudio-project/gsplat.git
   

5. In Blender preferences, set:
6. gsplat Examples Path: .../gsplat/examples (must contain simple_trainer.py)
7. gsplat Env: gsplat (or your env name)

Start Training in Blender

1. Open the Training panel (N key -> GS Capture -> Training)
2. Select a backend
3. Set Training Data Path (the capture output folder)
4. Set Training Output Path
5. Set iterations and optional extra arguments
6. Click Start Training