README.md

---
license: other
license_name: research-only
license_link: https://3dmatch.cs.princeton.edu/
task_categories:
  - other
tags:
  - point-cloud
  - 3d-registration
  - 3dmatch
  - geometric-features
  - fcgf
  - correspondence
pretty_name: FCGF Preprocessed 3DMatch
size_categories:
  - 1K<n<10K
---

# FCGF Preprocessed 3DMatch Dataset

Preprocessed **3DMatch** training data used by
[**FCGF: Fully Convolutional Geometric Features (ICCV 2019)**](https://github.com/chrischoy/FCGF).

Each `.npz` file stores a fragment point cloud, and the accompanying `.txt` files list
overlapping fragment pairs (with their overlap ratio) used to sample positive
correspondences during training. This is the exact data produced by
[`scripts/download_datasets.sh`](https://github.com/chrischoy/FCGF/blob/master/scripts/download_datasets.sh)
in the FCGF repository.

## Contents

```
threedmatch/
├── <scene>@seq-XX_YYY.npz          # 2189 fragment point clouds
└── <scene>@seq-XX-<overlap>.txt    #  401 overlapping-pair lists
```

- **2,590** files total (2,189 `.npz` + 401 `.txt`), ~8.2 GB.
- Scenes are drawn from the standard 3DMatch compilation: `7-scenes`, `sun3d`,
  `bundlefusion`, `rgbd-scenes-v2`, `analysis-by-synthesis`, etc.

## Usage

Download with the Hugging Face CLI:

```bash
hf download chrischoy/FCGF-3DMatch --repo-type dataset --local-dir ./data
# data/threedmatch/*.npz
```

or from Python:

```python
from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="chrischoy/FCGF-3DMatch",
    repo_type="dataset",
    local_dir="./data",
)
```

Then train FCGF:

```bash
python train.py --threed_match_dir ./data/threedmatch/
```

A single fragment file can be inspected with NumPy:

```python
import numpy as np

data = np.load("data/threedmatch/7-scenes-chess@seq-02_000.npz")
print(data.files)          # e.g. ['pcd', ...]
xyz = data["pcd"]          # (N, 3) point coordinates
```

## License & attribution

This is a redistribution of preprocessed data derived from the
[3DMatch benchmark](http://3dmatch.cs.princeton.edu/), which itself aggregates
several RGB-D datasets (SUN3D, 7-Scenes, BundleFusion, RGB-D Scenes v2, and others).
It is provided **for non-commercial research purposes only**. Please also comply with
the licenses of the original constituent datasets and cite 3DMatch. The FCGF source
code is released separately under the MIT License.

## Citation

If you use this data, please cite FCGF and 3DMatch:

```bibtex
@inproceedings{FCGF2019,
    author    = {Christopher Choy and Jaesik Park and Vladlen Koltun},
    title     = {Fully Convolutional Geometric Features},
    booktitle = {ICCV},
    year      = {2019},
}

@inproceedings{zeng20163dmatch,
    author    = {Andy Zeng and Shuran Song and Matthias Nie{\ss}ner and
                 Matthew Fisher and Jianxiong Xiao and Thomas Funkhouser},
    title     = {3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions},
    booktitle = {CVPR},
    year      = {2017},
}
```