---
license: cc-by-nc-4.0
language:
  - en
tags:
  - hand-tracking
  - pose-estimation
  - mediapipe
  - robotics
  - imitation-learning
  - motion-dataset
  - quality-vision
pretty_name: High-Quality Dexterous Hand Movements (Sample)
size_categories:
  - 1K<n<10K
---

# High-Quality Dexterous Hand Movements — sample

This Hugging Face dataset is a **non-commercial sample** export from the **Quality Vision Motion Dataset Engine**: **MediaPipe Hands** (21 landmarks) + optional **dexterous analytics** and hand **motion_intelligence**.

**This is a sample, not the full commercial pack.**  
For **larger commercial batches** (e.g. **~140k+ HQ frames**, multiple clips segmented for **specific dexterous motions**), see **[Dataset pricing](https://qvision.space/dataset-pricing)** and contact **info@qvision.space**.

## Files

This repo mirrors the `dataset/` folder layout used in production exports:

| Path | Description |
|------|-------------|
| `data.jsonl` | Merged HQ rows (one JSON object per line). |
| `per_video/` | Per-source `video_*.jsonl` and `rejected_*.jsonl`. |
| `features.json`, `global_stats.json` | Sequence / batch aggregates (includes dexterous histograms). |
| `manifest.json`, `bulk_combined_manifest.json` | Export metadata + post-processing flags. |
| `export_quality_report.json` | HQ threshold summary + acceptance rate definition. |
| `SCHEMA.md`, `ONEPAGER.md` | Human-readable schema + one-page snapshot for sharing. |

Raw source videos are **not** included in this repo (JSONL + metadata only).

## Usage

```python
from datasets import load_dataset

ds = load_dataset(
    "Alaaharoun/High-Quality_Dexterous_Hand_Movements_Dataset",
    data_files="data.jsonl",
    split="train",
)
```