Skip to content

API Reference

Complete API documentation for the Dataphy SDK, auto-generated from code docstrings.

Overview

The Dataphy SDK is organized into several key modules:

  • Dataset Management: Loading, managing, and augmenting robotics datasets
  • VisionPack: Augmentation pipeline with transforms and adapters
  • Data Sources: Adapters for various data platforms and formats
  • Visualization: 3D visualization tools powered by rerun.io
  • CLI: Command-line interface implementation
  • I/O Operations: Input/output utilities and manifest management

Quick Start

The main entry points for the API are:

Dataset Loading

from dataphy.dataset.registry import create_dataset_loader, DatasetFormat

# Auto-detect format (recommended)
loader = create_dataset_loader("./dataset")

# Explicit format
loader = create_dataset_loader("./dataset", DatasetFormat.LEROBOT)

Episode Augmentation

from dataphy.dataset.episode_augmentor import EpisodeAugmentor

augmentor = EpisodeAugmentor(loader)
augmentor.augment_episode(
    episode_id=0,
    config_file="config.yaml"
)

VisionPack Pipeline

from dataphy.visionpack.pipeline import build_pipeline

pipeline = build_pipeline("config.yaml", device="cuda")
augmented_batch = pipeline(batch)

Type Hints and Annotations

All public APIs include comprehensive type hints for better IDE support and code safety. Import common types:

from typing import Union, List, Optional, Dict, Any
from pathlib import Path

Error Handling

The SDK uses descriptive exception types:

  • ValueError: Invalid parameters or data
  • FileNotFoundError: Missing files or directories
  • RuntimeError: Unexpected runtime errors
  • Custom exceptions for specific error cases

Use the navigation on the left to explore specific modules and classes. Each module page includes:

  • Class and function signatures
  • Detailed parameter descriptions
  • Return value documentation
  • Usage examples
  • Source code links