README.md

---
license: apache-2.0
task_categories:
  - robotics
tags:
  - robotics
  - lerobot
  - so100
  - imitation-learning
  - arabic
---

# Baseer — serums dataset (`baseer_serums`)

LeRobot dataset for the [**Baseer**](https://github.com/fatma936-sudo/baseer) project — an
Arabic voice-controlled assistive SO-100 arm (Fanar hackathon, Theme 4).

Teleoperated demonstrations of grasping a **hair serum** or **face serum** from a vanity table
and placing it at a fixed delivery zone, at varied object positions.

## Contents
- **28 episodes** = 13 hair serum + 15 face serum
- Single front camera, **640×480 @ 30 fps**, LeRobot format
- 15,965 frames; 6-DoF state + action (SO-100, Feetech STS3215)
- Two language tasks:
  - `Pick up the hair serum and place it in the delivery zone`
  - `Pick up the face serum and place it in the delivery zone`

Quality-checked: firm gripper closes on every episode (travel 68–100), no calibration/wrist anomalies.

## Trained model
[`55CancriE/baseer-smolvla-serums`](https://huggingface.co/55CancriE/baseer-smolvla-serums)
— SmolVLA, 20k steps, final loss 0.012.

## Usage (LeRobot)
```python
from lerobot.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("55CancriE/baseer_serums")
print(ds.num_episodes, ds.num_frames, ds.fps)
```

*Project: [github.com/fatma936-sudo/baseer](https://github.com/fatma936-sudo/baseer) · [project page](https://fatma936-sudo.github.io/baseer/)*