boazkg/gs-pool-cl-splats

GS-Pool on CL-Splats

This dataset is built on CL-Splats by Ackermann et al. (dataset, project page, paper, ICCV 2025). Its photographs and camera solves are theirs; please cite their work (below) whenever you use this data.

Change annotations for the five real-world scenes of CL-Splats (cone, cupboard, desk, room, shelf), and the reconstructions, evidence and results of GS-Pool 1.3.1 (paper, code) on them: 29 annotated day-1 views, and 30 members (5 scenes, two trainers, seeds 22, 23, 24). CL-Splats' photographs are not included; the code downloads them from CL-Splats.

Structure

annotations/
├── <scene>/gt_mask/               change masks at the held-out day-1 views (PASLCD's convention)
├── <scene>/split.json             each day's training views and the held-out views
└── annotation_protocol.md, change_lists.json, review_log.csv, manifest.json
members/<scene>/Instance_1/
├── joint/<fastpgsr|vanilla>/seed<22|23|24>/
│   ├── t1/, t2/                   the two reconstructions (model.pt with the distilled field, model.ply)
│   ├── evidence/                  object pools, carrier scores, view statistics, and the per-photograph rows
│   │                              the view statistics are computed from (photo_trace)
│   └── paper_results/change_detection/   receipt.json, pool.json, decision trace, change masks
└── baselines/<mv3dcd|oscd>/change_detection/   MV-3DCD and O-SCD on the same masks: result.json, masks
inputs/<scene>/Instance_1/joint/
├── undistorted/                   the undistorted cameras (no photographs)
├── sam2/                          SAM2 label maps (automatic, linked)
├── dinov3/                        DINOv3 features (raw, pca)
└── repair/                        a repaired solve (room)

How to run

With the code installed, one member, starting from its released evidence:

bash run.sh clsplats vanilla cone 24 --from decision

The runner downloads only what that member needs. --from carriers and --from field start from the released reconstructions, --use-caches uses the released inputs/, and --data DIR points at a local copy of this dataset.

Citation

CL-Splats:

@inproceedings{ackermann2025clsplats,
  title     = {CL-Splats: Continual Learning of Gaussian Splatting with Local Optimization},
  author    = {Ackermann, Jan and Kulhanek, Jonas and Cai, Shengqu and Xu, Haofei and Pollefeys, Marc and Wetzstein, Gordon and Guibas, Leonidas J. and Peng, Songyou},
  booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year      = {2025}
}

GS-Pool:

@article{kerengil2026gspool,
  title   = {GS-Pool: Object-Level Change Detection in 3D Gaussian Splatting},
  author  = {Keren-Gil, Boaz and Gain, James and Marais, Patrick},
  journal = {arXiv preprint arXiv:2610.06688},
  year    = {2026}
}