LeRobot Dataset Tool and Visualizer is a web application for interactive exploration and visualization of robotics datasets, particularly those in the LeRobot format. It enables users to browse, view, and analyze episodes from large-scale robotics datasets, combining synchronized video playback with rich, interactive data graphs.

This tool is designed to help robotics researchers and practitioners quickly inspect and understand large, complex datasets. It fetches dataset metadata and episode data (including video and sensor/telemetry data), and provides a unified interface for:

• Navigating between organizations, datasets, and episodes
• Watching episode videos
• Exploring synchronized time-series data with interactive charts
• Analyzing action quality and identifying problematic episodes
• Visualizing robot poses in 3D using URDF models
• Paginating through large datasets efficiently

• Dataset & Episode Navigation: Quickly jump between organizations, datasets, and episodes using a sidebar and navigation controls.
• Synchronized Video & Data: Video playback is synchronized with interactive data graphs for detailed inspection of sensor and control signals.
• Overview Panel: At-a-glance summary of dataset metadata, camera info, and episode details.
• Statistics Panel: Dataset-level statistics including episode count, total recording time, frames-per-second, and an episode-length histogram.
• Action Insights Panel: Data-driven analysis tools to guide training configuration — includes autocorrelation, state-action alignment, speed distribution, and cross-episode variance heatmap.
• Filtering Panel: Identify and flag problematic episodes (low movement, jerky motion, outlier length) for removal. Exports flagged episode IDs as a ready-to-run LeRobot CLI command.
• 3D URDF Viewer: Visualize robot joint poses frame-by-frame in an interactive 3D scene, with end-effector trail rendering. Supports SO-100, SO-101, and OpenArm bimanual robots.
• Efficient Data Loading: Uses parquet and JSON loading for large dataset support, with pagination, chunking, and lazy-loaded panels for fast initial load.
• Responsive UI: Built with React, Next.js, and Tailwind CSS for a fast, modern user experience.

• Next.js (App Router)
• React
• Recharts (for data visualization)
• Three.js + @react-three/fiber + @react-three/drei (for 3D URDF visualization)
• urdf-loader (for parsing URDF robot models)
• hyparquet (for reading Parquet files)
• Tailwind CSS (styling)

This project uses Bun as its package manager. If you don't have it installed:

# Install Bun
curl -fsSL https://bun.sh/install | bash

Install dependencies:

bun install

Run the development server:

bun dev

Open http://localhost:3000 with your browser to see the result.

You can start editing the page by modifying src/app/page.tsx or other files in the src/ directory. The app supports hot-reloading for rapid development.

# Build for production
bun run build

# Start production server
bun start

# Run linter
bun run lint

# Format code
bun run format

• DATASET_URL: (optional) Base URL for dataset hosting (defaults to HuggingFace Datasets).

This application can be deployed using Docker with bun for optimal performance and self-contained builds.

docker build -t lerobot-visualizer .

docker run -p 7860:7860 lerobot-visualizer

The application will be available at http://localhost:7860.

docker run -p 7860:7860 -e DATASET_URL=your-url lerobot-visualizer

Contributions, bug reports, and feature requests are welcome! Please open an issue or submit a pull request.

The app was orignally created by @Mishig25 and taken from this PR #1055