# Masks for downsampling ScanNet++ V1 GS
The masks are for downsampling the Gaussian `.npy` files of `GaussianWorld/scannetpp_v1_mcmc_3dgs_preprocessed`.

The masks are created with [`/data/downsample_pc.py`](https://github.com/amhalacheva/GaussianPlan/blob/baseline_structured_output/data/downsample_pc.py)

1. Download `GaussianWorld/scannetpp_v1_mcmc_3dgs_preprocessed` in `/work/$USER` - only `train` and `val` split
2. Download the masks here (`amhalacheva/spp_v1_downsampled_gs`)
3. Apply the masks on the `.npy`s from scannetpp_v1_mcmc_3dgs_preprocessed
4. Rename `/work/$USER/scannetpp_v1_mcmc_3dgs_preprocessed` to `/work/$USER/scannetpp_v1_mcmc_3dgs_preprocessed_downsampled` as this is scan path in config for training.