# Technical Documentation: huggingface/lelab

> ℹ️ **Provenance:** Multi-Tier Ingestion · [CodeWiki Reference](https://codewiki.google/github.com/huggingface/lelab)

### Overview (`README.md`)

<h1 align="center">🦾 LeLab</h1>

<p align="center">
  <b>The official graphical interface for <a href="https://github.com/huggingface/lerobot">LeRobot</a>.</b>
</p>

<div align="center">

[](https://github.com/huggingface/leLab/blob/main/LICENSE)
[](https://huggingface.co/spaces/lerobot/LeLab)
[](https://discord.gg/q8Dzzpym3f)

</div>

**LeLab** is a web app that puts the full LeRobot workflow — calibrate, teleoperate, record, train, replay — into a single browser UI. Plug in your arm, open the app, and go. No CLI gymnastics, no keyboard prompts.

🤗 A web-native front door to LeRobot, designed so newcomers can get from "unboxing" to "training their first policy" in minutes.

🤗 Install and run everything with a single command.

## Quick Start

Grab the one-liner from the [LeLab Space](https://huggingface.co/spaces/lerobot/LeLab) — it installs and runs LeLab + LeRobot in a single command.

A page will automatically open in your browser and you are ready to go.

## What you can do

<div align="center">
  <table>
    <tr>
      <td>🎯 <b>Calibrate</b></td>
      <td>Guided web flow for both arms — no keyboard prompts.</td>
    </tr>
    <tr>
      <td>🕹️ <b>Teleoperate</b></td>
      <td>Move the leader, the follower mirrors it. Live joint streaming.</td>
    </tr>
    <tr>
      <td>📹 <b>Record</b></td>
      <td>Capture episodes into a LeRobotDataset, with cameras.</td>
    </tr>
    <tr>
      <td>🔎 <b>Browse</b></td>
      <td>Watch the episodes of a dataset on disk — every camera, frame by frame, with a joint-motion trace. No upload required.</td>
    </tr>
    <tr>
      <td>🧠 <b>Train</b></td>
      <td>Kick off a LeRobot training job, watch logs live.</td>
    </tr>
    <tr>
      <td>🤖 <b>Run inference</b></td>
      <td>Execute a trained policy on the follower.</td>
    </tr>
    <tr>
      <td>⏪ <b>Replay</b></td>
      <td>Re-run any recorded episode.</td>
    </tr>
    <tr>
      <td>☁️ <b>Upload</b></td>
      <td>Push your dataset to the <a href="https://huggingface.co/">Hugging Face Hub</a> in one click.</td>
    </tr>
  </table>
</div>

## Camera rotation

If a camera is mounted upside down or sideways, select its **Rotation** in the camera setup on the Calibration page.
Choose 0°, 90°, 180°, or 270° clockwise. For an upside-down wrist camera, choose 180°.

LeLab saves the rotation for each camera. Setup and teleoperation previews, recorded frames, and inference use the selected rotation.
Profiles without a rotation setting use 0°.

The resolution setting describes the camera's capture resolution. A 90° or 270° rotation swaps the output dimensions.
For example, a 640 × 480 capture becomes a 480 × 640 observation. The preview shows the complete rotated frame.

Before inference, select the same rotation used for the training dataset. Check it again if you change the camera binding.
Checkpoint image dimensions do not establish image orientation. For example, 0° and 180° have the same dimensions.
The setting applies to new observations; it does not change existing recordings or checkpoints.

## Resources

- **[LeRobot](https://github.com/huggingface/lerobot):** the underlying library — go here for everything beyond the UI.
- **[LeLab Space](https://huggingface.co/spaces/lerobot/LeLab):** try the UI in your browser.
- **[Discord](https://discord.gg/q8Dzzpym3f):** chat with the LeRobot community.
- **[CLAUDE.md](CLAUDE.md):** architecture rundown for contributors.

## Contribute

PRs welcome. Hot-reload mode for working on the code:

```bash
lelab --dev
```

Vite on `:8080`, uvicorn `--reload` on `:8000`. Installing LeLab builds the frontend bundle, so PRs never include `frontend/dist`.

<div align="center">
<sub>Originally hacked together by <a href="https://www.linkedin.com/posts/nicolas-rabault-_lerobot-hackathon-lerobot-ugcPost-7341065019368828930-jTnl/">Team LeLab at the 2025 LeRobot Worldwide Hackathon 🏆</a>, now maintained by the <a href="https://huggingface.co/lerobot">LeRobot</a> team at <a href="https://huggingface.co">Hugging Face</a> with ❤️</sub>
</div>

---

### Overview (`frontend/README.md`)

---
title: LeLab
emoji: ⚡
colorFrom: yellow
colorTo: red
sdk: docker
app_port: 7860
pinned: false
short_description: Simple Interface to use LeRobot
---

<img width="1283" alt="Screenshot 2025-06-19 at 00 23 55" src="https://github.com/user-attachments/assets/ce1b5888-d24b-4adf-a422-694d4b0aa751" />

# LeLab official Space repository

If you've used [LeLab](https://huggingface.co/spaces/lerobot/LeLab) and want to contribute or found a bug, this is the place to be. This repo is directly hooked up to the Hugging Face space.

Here's the equivalent [backend](https://github.com/huggingface/leLab) that keeps the FastAPI server you run to actually wrap the LeRobot library.

We'll be updating this README shortly.

---

### Overview (`frontend/public/so-101-urdf/README.md`)

# SO-ARM ROS2 URDF Package

A complete ROS2 package for the SO-ARM101 robotic arm with URDF description.

## 📋 Overview

This package provides a complete ROS2 implementation for the SO-ARM101 robotic arm, including:

- URDF robot description with visual and collision meshes
- RViz visualization with pre-configured displays
- Launch files for easy robot visualization
- Integration with MoveIt for motion planning
- Joint state publishers for interactive control

## 🎯 Original Source

https://github.com/TheRobotStudio/SO-ARM100/tree/main/Simulation/SO101

## 🚀 Key Improvements Made

### 1. **Complete ROS2 Package Structure**

- ✅ Proper `package.xml` with all necessary dependencies
- ✅ CMakeLists.txt for ROS2 build system
- ✅ Organized directory structure following ROS2 conventions

### 2. **Enhanced Visualization**

- ✅ Fixed mesh file paths for proper package integration

### Build Instructions

1. Clone this repository into your ROS2 workspace:

   ```bash
   cd ~/your_ros2_ws/src
   git clone <your-repo-url> so_arm_description
   ```

2. Build the package:
   ```bash
   cd ~/your_ros2_ws
   colcon build --packages-select so_arm_description
   source install/setup.bash
   ```

---

--- METRICS ---
- Files Extracted: 3
- Estimated Token Budget: ~1527 tokens
- Canonical Reference: https://codewiki.google/github.com/huggingface/lelab