Skip to main content

MPS Outputs - Hand Tracking

Project Aria's Machine Perception Services (MPS) uses Project Aria's SLAM (mono scene) camera images to estimate the hand movement of the wearer.

Hand Tracking outputs are available for all recordings made with SLAM cameras and can be requested using the MPS CLI or Desktop Companion app. When research partners submit data for processing, the data is only used to serve MPS requests. Partner data is not made available to Meta researchers or Meta’s affiliates. Go to MPS Data Lifecycle for more details about how partner data is processed and stored.

Go to MPS Code Snippets for how to:

  • Load the wrist and palm positions
  • Transform them to the world coordinate frame
  • Project them onto the SLAM or RGB images

Hand Tracking data can be visualized in Python or C++.

Hand Tracking Outputs

We currently provide Wrist and Palm Tracking data. The file outputs are:

  • wrist_and_palm_poses.csv - the coordinates of the wrist and palm positions in the device frame
  • summary.json - high-level report on MPS wrist and palm tracking

wrist_and_palm_poses.csv

wrist_and_palm_poses.csv contains the following fields:

ColumnTypeDescription
tracking_timestamp_usintTimestamp, in microseconds, of the SLAM camera frame in device time domain. This is the same time domain in which the MPS trajectory outputs are reported, so these timestamps can be directly used to infer the device pose from the MPS trajectory output.
left_tracking_confidencefloatA value between 0.0 and 1.0 indicating the confidence in the reported left wrist and palm positions. A value of -1.0 indicates that the left wrist and palm tracking data is missing for the frame, and the coordinates leftwrist_device and leftpalm_device should not be used.
tx_left_wrist_devicefloatX-coordinate of the left wrist position given in the device frame in meters.
ty_left_wrist_devicefloatY-coordinate of the left wrist position given in the device frame in meters.
tz_left_wrist_devicefloatZ-coordinate of the left wrist position given in the device frame in meters.
tx_left_palm_devicefloatX-coordinate of the left palm position given in the device frame in meters.
ty_left_palm_devicefloatY-coordinate of the left palm position given in the device frame in meters.
tz_left_palm_devicefloatZ-coordinate of the left palm position given in the device frame in meters.
right_tracking_confidencefloatA value between 0.0 and 1.0 indicating the confidence in the reported right wrist and palm positions. A value of -1.0 indicates that the right wrist and palm tracking data is missing for the frame, and the coordinates rightwrist_device and rightpalm_device should not be used.
tx_right_wrist_devicefloatX-coordinate of the right wrist position given in the device frame in meters.
ty_right_wrist_devicefloatY-coordinate of the right wrist position given in the device frame in meters.
tz_right_wrist_devicefloatZ-coordinate of the right wrist position given in the device frame in meters.
tx_right_palm_devicefloatX-coordinate of the right palm position given in the device frame in meters.
ty_right_palm_devicefloatY-coordinate of the right palm position given in the device frame in meters.
tz_right_palm_devicefloatZ-coordinate of the right palm position given in the device frame in meters.

The wrist and palm poses are given in the device frame in meters.

summary.json

summary.json contains the Operator Summary, listed in MPS Basics, as well as the following fields:

FieldTypeDescription
mean_confidencefloatAverage left_tracking_confidence and right_tracking_confidence value for frames with valid results
total_framesintTotal number of frames
valid_frame_fractionfloatFraction of frames that have reported a valid tracking result