binhpham/driving-dataset

---
license: apache-2.0
task_categories:
- robotics
tags:
- LeRobot
- rc-car
- self-driving
- imitation-learning
---

# RC Car Self-Driving Dataset

A LeRobot v3.0 dataset for training imitation learning policies on an RC car. Recorded via a remote operator interface over LiveKit, with a single front-facing camera and keyboard controls.

## Dataset Details

| Property | Value |
|---|---|
| Robot | RC car |
| FPS | 30 |
| Episodes | 10 |
| Total frames | 21,949 (~12 min) |
| Resolution | 320x180 |
| Task | `drive` |

## Features

- **observation.images.camera** — front-facing RGB video (320x180, h264)
- **observation.state** — previous frame's controls (6-dim float): `[forward, backward, left, right, speed, brake]`
- **action** — current controls (6-dim float): `[forward, backward, left, right, speed, brake]`

## Usage

```python
from lerobot.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset("binhpham/driving-dataset")
```

### Train with ACT policy

```bash
uv run lerobot-train --dataset binhpham/driving-dataset --policy act --steps 50000 --device mps
```

## Format

LeRobot v3.0 dataset format. See [LeRobot documentation](https://github.com/huggingface/lerobot) for details.