Skip to content

Dataphy SDK Documentation

A comprehensive Python SDK for robotics data I/O, dataset management, and augmentation packs.

Python 3.10+ License: MIT PyPI version

Features

  • Dataset Management: Fetch, load, inspect, and visualize robotics datasets
  • Episode Augmentation: Apply augmentations to specific episodes with camera selection
  • Dataset Augmentation: Create entirely new datasets with multiple augmented versions
  • Multi-Source Data: Support for Hugging Face Hub, LeRobot, and more
  • Interactive Visualization: 2D visualization powered by rerun.io
  • CLI Interface: Rich, user-friendly command-line interface
  • Backup/Restore System: Safe in-place augmentation with automatic backups
  • LeRobot Compatibility: Full support for v1.0, v1.1, v2.0, v2.1+ formats

Quick Start

Installation

# Install base dependencies
poetry install

# Install with visualization support
poetry install --extras rerun
poetry run dataphy-upgrade-rerun

# Verify installation
poetry run dataphy version

Basic Usage

# Fetch a LeRobot dataset
dataphy dataset fetch --format lerobot --repo-id carpit680/giraffe_clean_desk2 --output ./dataset

# Inspect dataset
dataphy dataset load --dataset-path ./dataset --info

# Augment single episode
dataphy augment dataset --dataset-path ./dataset --config aug.yaml --episode 0

# Create augmented dataset (3x larger)
dataphy augment dataset --dataset-path ./dataset --output-path ./augmented --config aug.yaml --num-augmented 2

# Visualize in 2D
dataphy dataset visualize --format lerobot --dataset-path ./dataset

Documentation Sections

Getting Started

Complete installation guide and first steps with the Dataphy SDK.

Tutorials

Step-by-step guides for common workflows:

API Reference

Complete API documentation auto-generated from code docstrings.

Tutorials

Step-by-step guides and code examples:

API Reference

Complete API documentation auto-generated from source code with type hints and examples.

Contributing

We welcome contributions! See our Contributing Guide for details.


Made with love by the Dataphy Team