README.md

---

pretty_name: ShapeNetPU
task_categories:
  - other
tags:
  - 3d
  - point-cloud
  - point-cloud-upsampling
  - shapenet
  - multi-view
  - geometry
size_categories:
  - 10K<n<100K
license: other
license_name: shapenet-terms-of-use
license_link: https://huggingface.co/spaces/ShapeNet/README/blob/main/TERMS.md
---


# ShapeNetPU

**ShapeNetPU** is a ShapeNet-derived dataset for **point cloud upsampling**, covering **13 object categories**. It provides paired sparse and dense point clouds, multi-view training images, and test-set reference meshes.

**Project / Code:** [ControlFlow3D](https://github.com/BaldFatTiger/ControlFlow3D) — **Code coming soon.**

**Paper:** [arXiv:2610.21190](https://arxiv.org/abs/2610.21190)

[Dataset summary](#dataset-summary) · [Download](#download) · [Directory structure](#directory-structure) · [Point groupings](#point-groupings) · [Point cloud format](#point-cloud-format) · [Citation](#citation-and-acknowledgments)

## Dataset Summary

| Property | Training split | Test split |
| :--- | ---: | ---: |
| Shapes | 30,395 | 5,367 |
| Object categories | 13 | 13 |
| Sparse input points per shape | 2,048 | 2,048 |
| Dense reference points per shape | 8,192 | 8,192 |
| Upsampling factor | 4× | 4× |
| Contiguous blocks per shape (derived) | 8 | 8 |
| Overlapping patches per PLY file (stored) | 10 | 10 |
| Rendered views per shape | 8 | Not included |
| Reference meshes | Not included | OBJ |

There are **35,762 shapes** in total. Each shape is identified by its ShapeNet category ID and model ID. The supplied training and test splits have no shared category/model IDs; no separate validation split is provided.

## Categories and Splits

| Synset ID | Category | Train | Test |
| :--- | :--- | ---: | ---: |
| `02691156` | airplane | 2,705 | 478 |
| `02828884` | bench | 1,154 | 204 |
| `02933112` | cabinet | 1,128 | 200 |
| `02958343` | car | 6,085 | 1,074 |
| `03001627` | chair | 4,116 | 727 |
| `03211117` | display | 899 | 159 |
| `03636649` | lamp | 857 | 151 |
| `03691459` | loudspeaker | 1,114 | 197 |
| `04090263` | rifle | 1,933 | 341 |
| `04256520` | sofa | 2,578 | 455 |
| `04379243` | table | 5,729 | 1,011 |
| `04401088` | telephone | 868 | 153 |
| `04530566` | watercraft | 1,229 | 217 |
| **Total** | **13 categories** | **30,395** | **5,367** |

## Download

Install the Hub client and the dependencies for the minimal reading example, then download the dataset:

```bash

pip install -U huggingface_hub numpy plyfile

hf download BaldFatTiger/ShapeNetPU --repo-type dataset --local-dir ShapeNetPU

```

For the test split only, append `--include "test/*"` to the download command. For reproducibility, pin a Hub commit with `--revision <commit>`.

## Directory Structure

```text

ShapeNetPU/

├── train/

│   └── <category_id>/

│       └── <model_id>/

│           ├── input/

│           │   ├── sparse_xyz.ply

│           │   └── rendering/

│           │       ├── <view_id>.png          # 8 selected views

│           │       ├── renderings.txt        # Ordered view filenames

│           │       └── rendering_metadata.txt

│           └── gt/

│               └── gt_xyz.ply

└── test/

    └── <category_id>/

        └── <model_id>/

            ├── input/

            │   └── sparse_xyz.ply

            └── gt/

                ├── gt_xyz.ply

                └── gt_mesh.obj

```

- `<category_id>` is a ShapeNet synset ID; `<model_id>` is the original model identifier.
- `input/sparse_xyz.ply` contains the sparse input cloud.
- `gt/gt_xyz.ply` contains the dense reference cloud.
- `input/rendering/` is included in the **training split only**.
- `gt/gt_mesh.obj` is included in the **test split only** and can be used for mesh-based evaluation, such as point-to-surface distance.

## Point Groupings

The same stored points support two complementary grouping views:

| Grouping | Groups per shape | Definition | Overlap | Points per group: sparse / dense |
| :--- | ---: | :--- | :--- | ---: |
| **Contiguous blocks** | 8 | Equal-sized consecutive ranges in the original vertex order | No | 256 / 1,024 |
| **Overlapping patches** | 10 | Groups specified by the stored `patch.vertex_indices` lists | Yes | 256 / 1,024 |

- **Blocks are implicit:** split the original vertex array into eight consecutive groups. Each point belongs to exactly one block; no separate block element is stored in the PLY.
- **Patches are explicit:** gather points using the zero-based index lists in the PLY `patch` element. Indices refer to vertices in the **same file**, and a point can belong to multiple patches.
- The loader names these modes `part` and `patch`, respectively. They are grouping conventions, **not semantic part labels**.

Preserve the original vertex order when constructing blocks. If vertices are reordered, carry their block memberships and remap patch indices to the new vertex positions.

## Point Cloud Format

Point clouds use **binary little-endian PLY** with the following fields:

| Element | Property | Meaning |
| :--- | :--- | :--- |
| `vertex` | `x`, `y`, `z` | Float32 point coordinates |
| `vertex` | `dense_index` | Per-file integer indexing metadata; not a sparse-to-GT correspondence map |
| `patch` | `center_x`, `center_y`, `center_z` | Float32 patch-center coordinates |
| `patch` | `vertex_indices` | List of zero-based indices into this file's vertex array |

The header comment `patch_index_space local_vertex` records this indexing convention.

### Minimal Reading Example

After downloading, this example reads one sparse cloud and constructs both grouping views:

```python

import numpy as np

from plyfile import PlyData



ply = PlyData.read("ShapeNetPU/test/02691156/10155655850468db78d106ce0a280f87/input/sparse_xyz.ply")

v = ply["vertex"].data

points = np.column_stack([v["x"], v["y"], v["z"]])

blocks = points.reshape(8, -1, 3)  # 8 contiguous, non-overlapping blocks

patches = [points[idx] for idx in ply["patch"].data["vertex_indices"]]

```

Use the corresponding `gt/gt_xyz.ply` file to read the dense reference cloud. Blocks have shape `(8, 256, 3)` for sparse inputs and `(8, 1024, 3)` for dense references; `patches` contains ten point arrays.

## Multi-View Renderings

Each training shape has **eight 137 × 137 RGBA PNG images**, selected from the source 24-view rendering set. Selected filenames range from `00.png` to `23.png` and can differ between shapes. Renderings are included in the training split only.

| File | Contents |
| :--- | :--- |
| `<view_id>.png` | A selected rendered view |
| `renderings.txt` | The eight selected image filenames, in view order |
| `rendering_metadata.txt` | Eight rows of camera parameters, aligned with `renderings.txt` |

The five metadata columns are **azimuth, elevation, in-plane rotation, distance, and field of view**, following the [upstream rendering convention](https://github.com/chrischoy/3D-R2N2/issues/12#issuecomment-339119416).

## Usage Notes

- Use the supplied train/test split. If validation data are needed, derive them from the training split.
- For point-cloud-only evaluation, use sparse test clouds as inputs; dense clouds and meshes are evaluation references.
- Apply consistent coordinate transforms to inputs, predictions, references, and meshes when computing distances, and report normalization and metric scaling.

## Citation and Acknowledgments

If you use ShapeNetPU in your research, please cite [ControlFlow3D](https://arxiv.org/abs/2610.21190):

```bibtex

@misc{liu2026controlflow3d,

  title={ControlFlow3D: Distilling Multi-View Knowledge into Latent Flow Matching for Point Cloud Upsampling},

  author={Yuang Liu and Zhi Zuo and Zhengkai Zhao and Lirui Zhang and Pan Gao and Hao Feng and Zhengzhe Liu},

  year={2026},

  eprint={2610.21190},

  archivePrefix={arXiv},

  primaryClass={cs.CV},

  url={https://arxiv.org/abs/2610.21190},

}

```

Please also cite ShapeNet, the source of the underlying 3D models:

```bibtex

@misc{chang2015shapenetinformationrich3dmodel,

      title={ShapeNet: An Information-Rich 3D Model Repository},

      author={Angel X. Chang and Thomas Funkhouser and Leonidas Guibas and Pat Hanrahan and Qixing Huang and Zimo Li and Silvio Savarese and Manolis Savva and Shuran Song and Hao Su and Jianxiong Xiao and Li Yi and Fisher Yu},

      year={2015},

      eprint={1512.03012},

      archivePrefix={arXiv},

      primaryClass={cs.GR},

      url={https://arxiv.org/abs/1512.03012},

}

```

We thank the ShapeNet contributors and the original 3D model creators for the source assets.

## License and Terms of Use

ShapeNetPU is derived from [ShapeNet](https://shapenet.org/). The underlying assets remain subject to the [ShapeNet Terms of Use](https://huggingface.co/spaces/ShapeNet/README/blob/main/TERMS.md), which restrict use to **non-commercial research and educational purposes**, and to any applicable licenses of the original model sources.

Users must review and comply with the upstream terms before using the data. ShapeNet's terms permit sharing with research associates and colleagues only if they first agree to be bound by those terms. This dataset card does not grant unrestricted redistribution rights or additional rights to the underlying third-party assets.

## Contact

For questions about ShapeNetPU or to report a data issue, please open an issue in the [ControlFlow3D repository](https://github.com/BaldFatTiger/ControlFlow3D/issues). Include the split, category ID, and model ID when reporting a sample-specific issue.