metadata
pretty_name: RLGS Release Assets
tags:
- 3d-gaussian-splatting
- robustnerf
- rlgs
RLGS release assets
This dataset repository hosts small, versioned auxiliary files needed to reproduce experiments from RLGS.
RL-AsymGS RobustNeRF masks
rl_asymgs/robustnerf_masks/ contains the precomputed multi-cue masks used by
the frozen 30,000-iteration RL-AsymGS runs on Yoda, Crab2, Statue, and Android.
The files correspond to the *-undistorted RobustNeRF scene layouts expected
by rl_asymgs/train.sh.
| Scene | File | SHA-256 |
|---|---|---|
| Yoda | multi_cue_masks_yoda-undistorted.npz |
f888c1b315819bb9a81fab6a9264681afd45624cfa8454ed70e1910ff06db298 |
| Crab2 | multi_cue_masks_crab2-undistorted.npz |
1163aca63654a351e321a4724e1a8a013d93c149ac9527b54cab1193c46d9ab3 |
| Statue | multi_cue_masks_statue-undistorted.npz |
9a2a8c307f5e18fd2cc4286612c75b66f6c447523f2e32012d4f552d82a3704a |
| Android | multi_cue_masks_android-undistorted.npz |
5b448093748c52dd75e242fffea9f8f4096e827145dc5a65ea2cd623347a27c8 |
The public RLGS downloader installs and verifies these files automatically:
cd rl_asymgs
bash scripts/download_dataset.sh --scene all
The original RobustNeRF images, COLMAP reconstructions, DINOv2 weights, and trained checkpoints are not redistributed here. Users should obtain those resources from their original providers and follow their respective terms.