README.md

---
license: mit
language:
  - en
tags:
  - 3d
  - point-cloud
  - mesh
  - classification
  - modelnet
pretty_name: ModelNet10
size_categories:
  - 1G<n<10G
---

# ModelNet10

[ModelNet10](https://modelnet.cs.princeton.edu/) is a subset of ModelNet with **10 object categories** of 3D CAD models, widely used for shape classification and point-cloud experiments. This repository packages the meshes as **OFF** files with a CSV index for use with the [Hugging Face `datasets`](https://huggingface.co/docs/datasets) library.

## Dataset structure

- `metadata_modelnet10.csv` — index with columns: `object_id`, `class`, `split`, `object_path`
- `ModelNet10/` — mesh files (`.off`) organized by category and split
- `modelnet10.py` — `GeneratorBasedBuilder` loading script

Each example exposes string fields `object_id`, `class`, `split`, and `file_path` (absolute path to the `.off` file after resolution).

## Usage

Load from the Hub (downloads and caches on first use):

```python
from datasets import load_dataset

ds = load_dataset("naderalfares/ModelNet10")
```

Load from a local directory that contains `metadata_modelnet10.csv` and the `ModelNet10/` folder:

```python
from datasets import load_dataset

ds = load_dataset(
    "path/to/modelnet10.py",
    name="default",
    data_dir="path/to/ModelNet10",
)
```

Access splits:

```python
train = ds["train"]
row = train[0]
print(row["class"], row["file_path"])
```