Aria Gen2 Pilot Dataset

A Python package for loading, processing, and visualizing data from the Aria Gen2 Pilot Dataset. This dataset contains multimodal sensor data from Project Aria Gen2 glasses, including raw sensor streams, real-time machine perception outputs, and post-processed algorithm results.

Features

• VRS Data Loading: Access raw sensor data (RGB/SLAM cameras, IMU, audio, GPS, etc.)
• Machine Perception Services (MPS): Load SLAM trajectories and hand tracking results
• Algorithm Outputs: Process heart rate monitoring, diarization, hand-object interaction, egocentric voxel lifting, and stereo depth data
• Visualization Tools: Built-in viewer with Rerun integration for 3D visualization
• Unified API: Single AriaGen2PilotDataProvider interface for all data types

Installation

# Deactivate conda if active
conda deactivate

# Remove existing environment if it exists
rm -rf ~/projectaria_gen2_python_env

# Create new Python virtual environment
python3.12 -m venv ~/projectaria_gen2_python_env

# Activate the environment
source ~/projectaria_gen2_python_env/bin/activate

# Upgrade pip
python3 -m pip install --upgrade pip

# Install the package with all dependencies
python3 -m pip install projectaria-gen2-pilot-dataset'[all]'

Tutorials

Get started with these comprehensive tutorials:

• VRS Data Loading - Loading raw sensor data
• MPS Data Loading - Loading Machine Perception Services data
• Algorithm Data Loading - Loading algorithm output data
• Multiple Sequences Synchronization - Synchronizing data between multiple devices

Dataset Structure

Each sequence contains:

• Raw VRS files: Sensor streams from Aria Gen2 devices
• MPS outputs: SLAM trajectories and hand tracking results
• Algorithm data: Heart rate, diarization, hand-object interaction, depth estimation, and 3D scene reconstruction

Visualization

Launch the interactive viewer:

aria_gen2_pilot_dataset_viewer --sequence-path /path/to/sequence

License

This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). See the LICENSE file for details.