Style-Splat: Official Code Repository

🚀 Quick Start

Prerequisites

• CUDA-compatible GPU
• Conda package manager

Installation

1. Clone the repository:

   git clone https://github.com/bernard0047/style-splat --recursive
   

2. Set up the Conda environment:

    conda create -n style-splat python=3.8 -y
    conda activate style-splat
   
    conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
    pip install plyfile==0.8.1 tqdm scipy wandb opencv-python scikit-learn lpips
   
    pip install submodules/diff-gaussian-rasterization
    pip install submodules/simple-knn
   

3. Install DEVA tracking:

    cd style-splat
    git clone https://github.com/hkchengrex/Tracking-Anything-with-DEVA.git
    cd Tracking-Anything-with-DEVA
    pip install -e .
    bash scripts/download_models.sh
    cd ../..
   

🎨 Usage

1. Preparing Pseudo Labels

Before training or applying style transfer, you need to prepare pseudo labels for your dataset.

1. Edit scripts/prepare_pseudo_label.sh:

   • Set the DATASET_ROOT variable to your dataset's root folder
   • Adjust the SCALE variable as needed for your dataset

   Example:

   DATASET_ROOT="/path/to/your/dataset"
   SCALE=2
   

2. Generate object masks:

    bash scripts/prepare_pseudo_label.sh

   • This will create an object_mask folder in your dataset directory.

2. Training Gaussian Grouping Instance

• Train the model to group Gaussians:

  python train.py -s data/your_scene -r 2 \
  --config config/gaussian_dataset/train.json --port 8000 \
  -m outputs/your_scene_output

3. Style Transfer

1. Create a configuration file:
   • Copy config/object_style_transfer/bear.json for your scene
   • Use Image Color Picker to select object IDs
   • Update select_obj_id in your configuration file
2. Run style transfer:
   • Use edit_object_style_transfer.py for single object
   • Use edit_object_style_transfer_multiple.py for multiple objects
   • See transfer_style.sh for example usage

📚 Additional Resources

• DEVA Tracking
• Image Color Picker: An online tool to help identify object IDs from your mask images.

✨ Acknowledgements

We would like to express our gratitude to the following projects that have significantly contributed to the development of Style-Splat:

• 3D Gaussian Splatting for Real-Time Radiance Field Rendering
• Gaussian-Grouping

📄 License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

---