README.md

---
license: cc-by-4.0
dataset_info:
  features:
    - name: data
      dtype: float32
      shape: [null, 2, 135]   # variable length sequences
    - name: label
      dtype: int64
      description: "0–117 (see MathClassroom_ASL_labels.json)"
  splits:
    train: 9,843
    validation: 1,227
    test: 1,287
---

# PrimaryMath-Classroom-ASL-118  
**The first public ASL keypoint dataset designed specifically for primary-school mathematics & classroom communication**

**118 essential signs** • 12,357 clean Mediapipe Holistic sequences • Ready-to-train splits

Perfect for real-time mobile apps, classroom translation tools, and educational technology for deaf children.

### Included signs (examples)
`zero–twenty`, `thirty`, `forty`, `fifty`, `hundred`, `plus`, `minus`, `equal`, `multiply`, `divide`, `book`, `paper`, `pencil`, `teacher`, `student`, `hello`, `yes`, `no`, `help`, `again`, `please`, `thank you`, `sorry`, `red`, `blue`, `green`, `circle`, `square`, `triangle`, `school`, `how many`, `more`, `less`, `count`, `good`, `table`, `chair`, `listen`, `look`, `stand up`, `sit down`, ...

### Stats
| Split       | Samples |
|-------------|---------|
| Train       | 9,843   |
| Validation  | 1,227   |
| Test        | 1,287   |

Keypoints: Mediapipe Holistic (135 landmarks × 2 coordinates)

### Citation (original sources — MUST CITE)
```bibtex
@inproceedings{lazo2025wholebodypose,
  title={WholeBodyPose: A Unified End-to-End Framework for Sign Language Recognition and Pose-Based Training Data},
  author={Lazo-Quispe, Cristian and Castro-Cruz, Renato and Salazar-Espinosa, Mauricio and Bejarano, Gissella},
  booktitle={International Conference on Machine Vision (ICMV)},
  year={2025},
  organization={SPIE}
}
Also based on WLASL, MSASL, and ASLCitizen datasets.
Curated by
Umama Saif — Final Year Project 2025
Made with love for deaf education and accessibility ❤️
Feel free to use, improve, and deploy in real classrooms!