Skip to main content

Project Aria Research Kit Release Notes

Writing Sample

This page is based on technical writing I did for Project Aria. Some of the links will be broken, as I did not bring across the whole site. For the most up to date documentation go to Project Aria Docs.

Evolving the SW Release Notes

Drawing on my experience creating SW Release Notes for Halo Infinite, I introduced much more robust SW Release Notes for Project Aria. I created a consistent structure and championed adding Known Issues and Bug Fixes to our release notes. Over time, engineers came to see that communicating with our users around Known Issues and Bug Fixes helped us demonstrate that we were listening, provided transparency around our practices, and provided valuable information to User Support.

These SW Release notes are for the Aria Research Kit (ARK), and are for Meta SW that supports Academic Research Partners using Project Aria glasses. To find out how to become a research partner, go to projectaria.com.

Our Open Source Project Aria Tools SW Release notes are posted to Project Aria Tools GitHub repository, using Tags > Releases.

July 2024​

Mobile Companion App v170 and Time Synchronized Recordings Between Multiple Project Aria Devices​

Dear Academic Partners,

Mobile Companion app V170 and Aria OS 4970205.890.70 are now available. With the release of Aria OS 4970205.890.70, you can now create time synchronized recordings between multiple devices via the Client SDK.

  • Recent documentation improvements include:
    • Renamed The Project Aria Tools documentation site has been renamed to reflect its broader scope, it’s now called Project Aria Docs. The Data Utilities section has been renamed to Project Aria Tools.
    • Updated Intro page and About ARK
    • New Project Aria FAQ

Aria Client SDK (TICSync support and general improvements)​

Users can now use the Aria Client SDK to capture time-synchronized data between multiple Project Aria glasses (and potentially other devices) on the one Wi-Fi network via TICSync. Using the Aria hotspot feature, one pair of glasses (server) acts as a Wi-Fi access point that forms a network between all glasses. Time requests from all other glasses go to the server/leader device, creating a synchronized time reference.

To use TICSync you’ll need to:

In addition:

  • The Aria Client SDK is now supported on Python 3.12
  • Wi-Fi connections to Aria glasses have been made more reliable by enabling sticky connection

Known Issues​

  • The Client SDK is a software developer kit. While we have created code samples to enable a range of features, we have not created fully fledged applications. We encourage people to make this code their own, and to use the code samples as a starting point.

Bug Fixes​

  • Information in error messages has been improved
  • Reliability of multiple Project Aria devices connected over USB has been improved

Mobile Companion app​

The Aria Mobile App v170 is now available. Follow the in-app prompt or go to the Downloads page to get the latest version for Android. For iOS, select Update in the TestFlight app. Please update the app to access the latest bug fixes and finalized retheming of all app pages.

Project Aria Latest OS​

The latest Project Aria OS build 4970205.890.70 (June 28, 2024).

Project Aria glasses OS should automatically update when plugged into power and connected to Wi-Fi. Go to How to Update Your Aria Glasses’ OS if you want to manually update your glasses.

Questions, Concerns, Feedback?​

If you encounter any problems using Project Aria support, please contact us. There are several ways for getting user support.

We encourage you to post issues to GitHub, so that they can be tracked.

If you are a Research Partner with access to the Aria Research Kit there are several other support options available, although GitHub is ideal for bug reports and feature requests.



May 2024​

Aria Mobile App v165 is now available​

Dear Academic Partners,

The Aria Mobile App v165 is now available. Go to the Downloads page to get the latest version for Android. For iOS, select Update in the TestFlight app. Please update the app to access the following features and bug fixes.

Project Aria recently released Project Aria Eye Tracking, an open source inference code for the Pre March 2024 Eye Gaze Model used by MPS. This model:

  • Provides the ability to generate eye gaze data using the old model, if that is more suited to your research use case or if you’re not able to request MPS
  • This code can be used on downloaded data or when streaming data using the Aria Client SDK

Bug Fixes​

Various bug fixes to improve the overall usage.

Project Aria Latest OS​

The latest Project Aria OS build 4968951.840.70 (April 25th, 2024).

Questions, Concerns, Feedback?​

Go to our Support page for ways to get in touch.


April 2024​

Announcing: New Custom Recording Profile Capabilities​

We’re pleased to announce that Aria Mobile Companion app v160 contains the ability to configure individual sensors when creating custom recordings using the Mobile Companion app.

How to access the update

You’ll be prompted to update the app when you open it, and you can also manually download it.

You’ll need Aria OS version 4968999.0.70 (April 25th, 2024) or later to be able to set custom recording profiles.

Custom Recording Profile​

To keep the new Custom Recording Profile feature approachable, and to reduce our testing matrix, we are launching this feature with a limited set of parameters. If people find the feature useful, we are open to unblocking research needs by providing the ability to include additional parameters. Please use one of our support channels to make feature requests.

To start a recording with a custom recording profile:

  1. In the Aria app Dashboard, select New Recording Session
  2. Select Recording Profile
  3. The first recording profile in the list will be Custom Profile
    • You may need to scroll up to see this option
  4. Select the Custom Profile
  5. Select Edit parameters
  6. Adjust sensor settings
    • Certain combinations, such as RGB Cameras at 2880 x 2880 at 15 or more fps will trigger warnings as they cause elevated thermal loads and reduced battery life
    • Tap More info on the Recording Profiles Tab to see more details, such as auto exposure

Bug Fixes​

  • An issue where the More info link on the top right of the Recording Profiles page was difficult to see has been resolved.

Known issues​

  • Custom Recordings is a new feature! If you encounter data quality issues, please help us track down the issue by emailing the repro steps (including sensor configuration) to AriaOps@meta.com.

Questions, Concerns, Feedback?​

Go to our Support page for ways to get in touch.


March 2024​

Announcing New MPS, Recording Profiles, and Updates to the Companion Apps​

Dear Academic Partners,

This release includes new Machine Perception Services (MPS) capabilities, Wrist and Palm Tracking data and a revised Eye Gaze model that provides depth estimations, as well as four new recording profiles, and an Automatic Speech Recognition speech to text code sample.

How to access the new features​

  • MPS
    • NEW SERVICE: Hand Tracking
      • Request Hand Tracking via the MPS CLI or Desktop app to get Wrist and Palm Tracking data!
    • UPDATED SERVICE: The new Eye Gaze model includes depth estimations
      • Backwards compatibility with Project Aria Tools will be maintained for data generated using the old model
    • Use the MPS CLI to do bulk requests on existing data
      • Raw uploaded data will now be stored for 30 days for MPS processing and reprocessing
  • Desktop app v38 [MacOS and Windows]
    • Download the latest version of the app for the ability to request Hand Tracking MPS via Desktop app, and to get bug fixes
  • Project Aria Tools v1.5
    • New code sample that generates speech to text outputs from VRS files
    • Install via Pip to be able to request Hand Tracking data using the MPS CLI
  • Aria glasses OS 4968297.250.70
    • Update to the latest OS for four new recording profiles and essential bug fixes
    • Please make sure your glasses have been updated in the last two months
  • Aria Mobile Companion App v155
    • Download the latest version of the app to enjoy the updated look and bug fixes
  • Documentation

MPS Updates​

Once you upload data using the MPS CLI, it will be stored for 30 days for additional processing/reprocessing, instead of 24 hours.

New Hand Tracking MPS, Wrist and Palm Tracking​

For this new service, we use Aria's SLAM camera images to estimate the hand movements of the wearer. The wrist and palm poses are given in the device frame in meters. These outputs are automatically generated when doing the default MPS CLI request (or request hand_tracking mode) and are listed as Hand Tracking in the Desktop App.

Install Project Aria Tools vs 1.5 using Pip or install Desktop App v38 to be able to request this new service.

Update to Eye Gaze outputs - depth estimations​

We have introduced a new model for computing eye gaze, which produces left and right eye gaze directions along with the depth at which these gaze directions intersect.

This new model’s output schema adds to general_eye_gaze.csv and personalized_eye_gaze.csv in a way that ensures backward compatibility with Project Aria Tools.

Known Issues​

  • The new Eye Gaze model does not generate yaw_rads_cpf outputs
    • This value can be computed using helper functions and is automatically read when parsed with our data utilities.

Bug Fixes​

Minor bug fixes

Desktop Companion App v38​

Download v38 on Windows or MacOS to request Hand Tracking MPS and to get bug fixes.

MPS requests using the Desktop app have been slightly restructured, you no longer need to request Point Cloud data as an additional service. The request options are:

  • SLAM (6DoF Trajectory, Semi-Dense Point Cloud, Online Sensor Calibration)
  • Eye Gaze (General, plus personalized if the VRS file contains In-Session Eye Gaze Calibration)
  • Hand Tracking (Wrist and Palm Tracking)

The Streaming button in the dashboard has been renamed to Preview, to better reflect the capability provided by the Desktop app. Use the Client SDK with CLI to stream data.

Desktop app logs are now stored in ~/.aria/logs/aria_desktop_app_{date}_{time}.log

  • Please note, the streaming preview available through the Desktop app is optimized for Profile 12.

Known Issues​

  • If you encounter the file too small error when requesting MPS, it may be because the Desktop app is in a bad state. Restarting the app should resolve this issue
  • Selecting Cancel does not work when downloading MPS is in progress
  • [MacOS] When the app is launched with logging enabled, the Aria icon might not appear in the taskbar, but functions should not be impacted
  • [Windows] The Aria icon does not render correctly when pinned to the taskbar

Bug Fixes​

  • If starting recording/streaming fails the UI will now refresh
  • Issue where you can’t request MPS unless you interact with tick boxes has been resolved
  • Issue where the Desktop app did not quit completely on MacOS has been resolved

Project Aria Tools v1.5​

Project Aria Tools provides open tooling for anyone wanting to work with Aria data. For this release it enables users to request Hand Tracking data using the MPS CLI and provides a new ASR Python code sample.

Update Project Aria Tools:

Go to https://github.com/facebookresearch/projectaria_tools/releases/tag/1.5.0 for the full release log.

Speech to Text Code Sample (Python)​

This new ASR Python code sample shows how to use Faster Whisper to run Whisper Speech Recognition on an Aria audio stream. The ASR outputs are time aligned with Aria Device Time.

Go to the Automated Speech Recognition Readme for how to get started.

Speech to Text Outputs

Aria Glasses OS​

The latest Project Aria OS build 4968297.250.70 was released on March 11, 2024.

New recording profiles​

We’ve released the following experimental recording profiles. If you use a new profile, we’d love to hear about how it works for you.

Recording Profile 26: Minimal sensors, low RGB frame rate for extended recording times. The aim is 5+ recording hours while having RGB camera data at a low frame rate (unlike Recording Profile 20, which has IMU, but no RGB data).

Recording Profile 27: Designed for streaming over USB.

**Recording Profile 28: **Very high Eye Tracking camera frame rate.

Recording Profile 29: Low RGB frame rate, JPEG with decimation. This recording profile is the same as Profile 26, however 9 out of every 10 JPEG frames are skipped (decimation/sub-sampling). This is our first recording profile that uses decimation. The data from the RGB data stream is saved at a rate that archives a frame rate of 0.1FPS. The aim of this profile is to be able to record for 15+ hours.

Update to the latest version of the OS to access these profiles. Your glasses’ OS should automatically update when connected to Wi-Fi and power. Go to Updating OS to manually trigger an update.

You can see more details about recording profiles in the Mobile Companion app:

  1. Open the Mobile Companion App and select New Recording Session
  2. Select the existing recording profile (or Choose Profile if you have not set a default recording profile)
  3. Tap More Info in the top right corner
    • The More info link may be difficult to see, this will be fixed in a future version of the Mobile Companion app

Or go to Recording Profiles documentation in Project Aria Tools.

Known Issues​

  • The new recording profiles are experimental and may be tuned further in the future.

Bug Fixes​

  • Make sure your glasses’ OS is less than 2 months old, to make sure you have a fix for an issue that may cause sluggish behavior in your Aria glasses, difficulties updating the OS, streaming data and reduced upload times.
    • Updating OS includes troubleshooting steps if you have difficulties updating the OS.

Mobile Companion App v155​

Please keep in mind, you can’t initiate recordings through the Mobile app if your glasses’ OS is more than 2 months old. Please make sure you update your glasses’ OS.

Updated look and feel of the Mobile app​

The Mobile Companion app has a fresh new look! As part of this:

  • The Settings tab has been renamed Profile
  • Device Settings can be opened by tapping the glasses icon under Paired Glasses
Device Settings page from the Dashboard

Known Issues​

  • The More info link on the top right of the Recording Profiles page may be difficult to see.

Bug Fixes​

Minor bug fixes

Questions, Concerns, Feedback?​

Go to our Support page for ways to get in touch.


Feburary 2024​

Announcing Project Aria MPS CLI Availability and New MPS Feature​

Dear Academic Partners,

We are pleased to announce Project Aria MPS CLI and new MPS capability multi-trajectory map alignment! As part of our ongoing commitment to empowering researchers, the Aria MPS CLI allows you to submit jobs to Machine Perception Services programmatically. This will hopefully bring a new level of efficiency and flexibility to your workflow.

⭐️Key features for MPS CLI​

  • Request the following MPS services
    • Single and multi-sequence SLAM
      • SLAM Open/Closed loop trajectories
      • Semi-Dense point clouds and observations
      • Online sensor calibration
    • Eye gaze
      • General
      • Personalized
  • Recording Upload
    • VrsHealthCheck runs automatically before uploading. Upload is skipped if there is a failure.
    • Unfinished uploads can be resumed for up to 24 hrs
  • Processing
    • Improved efficiency from concurrent processing and upload of multiple recordings
    • Automatic download of outputs for immediate access
    • Enables re-processing of data within a set time-frame without the need for re-upload
  • Workflow integration Support
    • Run MPS CLI as part of a script of other workflow to run MPS on recordings via automation

How to get started with CLI​

To experience the power of Aria MPS CLI, go to the Project Aria MPS CLI page for how to use it.

Videos of this feature are available in the Project Aria Academic Partner Announcements, Feedback & Support workplace group

⭐️MPS: MULTI-SLAM/MULTI-SEQUENCE MAP ALIGNMENT​

This addition brings the ability to generate SLAM output in a common frame of reference and support for multi-sequence SLAM outputs in a common frame of reference. Please go to Multi-SLAM page for more information about the feature and go to the MPS CLI page for how to access it.

⭐️Feedback and Support​

If you encounter any problems using Aria MPS CLI, there are several ways to get user support:

  • Post to Project Aria Academic Partner Announcements, Feedback & Support workplace group
  • Post to Project Aria Discord
  • Email AriaOps@meta.com and you can always request a meeting through this email if you need higher bandwidth discussion.
  • Additional resources are available at Project Aria Github.