Python package to compress Gaussian Splats with Self-Organizing Gaussians
Code forked from gsplat's png_compression module and produces a compressed bundle suitable for rendering with PlayCanvas' SuperSplat.
Requires torch, torchpq (which requires cupy, and PLAS, which require CUDA. These must be manually installed as they require installation against a specific version of CUDA (the one you have installed).
For instance, if you're running CUDA 12.6 on Windows you may install these dependencies (ideally in some kind of virtual environment):
pip install torch --index-url https://download.pytorch.org/whl/cu126
pip install cupy-cuda12x
pip install torchpq
pip install git+https://github.com/fraunhoferhhi/PLAS.git
pip install sogs
sogs-compress --input-ply your_ply_file.ply --output-dir directory_to_store_images_and_metadata