schwein69/hagrid-subset

---
license: mit
task_categories:
- image-classification
- object-detection
pretty_name: HaGRID Gesture Recognition Subset
size_categories:
- 10K<n<100K
tags:
- gesture-recognition
- computer-vision
- hand-gestures
---

# HaGRID Gesture Recognition Subset

## Dataset Description

A curated subset of the HaGRID (Hand Gesture Recognition Image Dataset) containing 24 gesture classes for training gesture recognition models.

### Dataset Summary

- **Total Images:** 19,200
- **Gesture Classes:** 24
- **Samples per Class:** 800
- **Image Format:** JPEG
- **Average Image Size:** ~302 KB

### Splits

| Split | Images | Percentage |
| ----- | ------ | ---------- |
| Train | 14,592 | 76% |
| Val | 1,728 | 9% |
| Test | 2,880 | 15% |

### Gesture Classes

1. call
2. dislike
3. fist
4. four
5. grabbing
6. grip
7. like
8. middle_finger
9. mute
10. no_gesture
11. ok
12. one
13. palm
14. peace
15. peace_inverted
16. point
17. rock
18. stop
19. stop_inverted
20. three
21. three2
22. three3
23. two_up
24. two_up_inverted

## Dataset Structure

```
data/
├── train/
│ ├── call/ (608 images)
│ ├── dislike/ (608 images)
│ └── ...
├── val/
│ └── ...
└── test/
└── ...
annotations.csv (19,200 entries)
```

## Usage

### Loading with Hugging Face Datasets

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("YOUR_USERNAME/hagrid-subset")

# Access splits
train_data = dataset["train"]
val_data = dataset["validation"]
test_data = dataset["test"]
```

### Loading with Pandas

```python
import pandas as pd
from PIL import Image

df = pd.read_csv("hf://datasets/YOUR_USERNAME/hagrid-subset/annotations.csv")
train_df = df[df['split'] == 'train']
```

## Citation

If you use this dataset, please cite the original HaGRID dataset:

```bibtex
@inproceedings{hagrid2022,
title={HaGRID--HAnd Gesture Recognition Image Dataset},
author={Kapitanov, Alexander and Kvanchiani, Karina and Nagaev, Alexander and Kraynov, Andrey and Makhliarchuk, Maxim},
booktitle={2022 International Conference on Robotics and Artificial Intelligence (ICRAI)},
year={2022}
}
```

## Dataset Creation

This subset was created using stratified sampling to maintain train/val/test ratios while ensuring class balance.

**Validation Status:** ✅ Fully validated with 100% annotation coverage

## License

MIT License (inherited from original HaGRID dataset)