README.md

---
license: cc-by-nc-4.0
tags:
  - 3d-gaussian-splatting
  - panorama
  - equirectangular
  - 3d-reconstruction
  - gaussian-splatting
---

<div align="center">

# GGPS — Datasets & Pretrained Models

**Geometry and Gradient-based Partitioning for Panoramic Outdoor Reconstruction**

[![Code](https://img.shields.io/badge/GitHub-Code-181717?logo=github&logoColor=white)](https://github.com/Insta360-Research-Team/GGPS)
[![arXiv](https://img.shields.io/badge/arXiv-Paper-b31b1b?logo=arxiv&logoColor=white)](https://arxiv.org/abs/2607.08769)
[![Homepage](https://img.shields.io/badge/Project-Homepage-1a73e8?logo=googlechrome&logoColor=white)](https://insta360-research-team.github.io/GGPS-Website/)

</div>

## 📦 Overview

This repository hosts the **datasets** and **pretrained `.ply` Gaussian models** for
PanoLOG / GGPS — panoramic (equirectangular / ERP) outdoor 3D Gaussian Splatting
reconstruction. For training and rendering code, see the
[GGPS code repository](https://github.com/Insta360-Research-Team/GGPS).

- `datasets/` — per-scene capture data (ERP panoramas + openMVG reconstruction), one archive per scene.
- `ply/` — pretrained / reconstructed Gaussian Splatting `.ply` models.

## 🔢 Dataset overview

| Scene | Archive | Panoramas | Reconstruction |
|-------|---------|----------:|----------------|
| FTP   | `datasets/FTP.zip` | 354  | openMVG (`sfm_data.bin`/`.json`, `colorized.ply`, `cloud_and_poses.ply`) |
| NSC   | `datasets/NSC.zip` | 1862 | openMVG (same as above) |
| NSK   | `datasets/NSK.zip` | 576  | openMVG (same as above) |
| **Total** | | **2792** | |

All panoramas are equirectangular (ERP) `.png`.

## 🗂️ Layout

```text
GGPS/
├── datasets/            # per-scene archives (FTP.zip, NSC.zip, NSK.zip, ...)
│   └── README.md
└── ply/                 # pretrained .ply Gaussian models
    └── README.md
```

Each dataset archive expands to the openMVG → COLMAP layout expected by the code repo:

```text
<scene>/
├── images/                  # ERP panoramas
└── reconstruction/
    ├── sfm_data.bin         # openMVG binary (sfm_data.json can be re-exported from this)
    ├── sfm_data.json        # openMVG JSON
    ├── colorized.ply        # colorized sparse point cloud (3DGS init)
    └── cloud_and_poses.ply
```

## ⬇️ Download

With the Hugging Face CLI:

```bash
pip install -U "huggingface_hub[cli]"

# whole repo
hf download Insta360-Research/GGPS --repo-type model --local-dir GGPS

# a single scene archive
hf download Insta360-Research/GGPS datasets/FTP.zip --repo-type model --local-dir .
```

## 📄 License

Released under **CC BY-NC 4.0** (non-commercial use only), consistent with the GGPS code repository.

## 📌 Citation

```bibtex
@article{panolog2026,
  title   = {Geometry and Gradient-based Partitioning for Panoramic Outdoor Reconstruction},
  author  = {Chen, Weijian and Yao, Weibo and Zhang, Yuhang and Tang, Xiaolin and
             Wang, Guo and Zhang, Weijun and Gao, Xitong and Chen, Yihao and
             Qin, Hongde and Qi, Lu},
  journal = {arXiv preprint arXiv:2607.08769},
  year    = {2026}
}
```