Skip to main content

Calibration Code Snippets

This section covers how to use Project Aria Tools to:

Go to the Project Aria FAQ for more calibration information and resources.

Accessing device calibration​

Device calibration stores:

  • The device's CAD model, which contains the 6DoF poses of sensors of the device as designed.
  • The calibration of all sensors on a single Aria device. See the Accessing sensor calibration section for details.
  • The device's sub-type (DVT-S or DVT-L to indicate small or large)
from projectaria_tools.core import data_provider, calibration
from projectaria_tools.core.stream_id import StreamId

vrsfile = "example.vrs"
provider = data_provider.create_vrs_data_provider(vrsfile)

# returns None if vrs does not have a calibration
device_calib = provider.get_device_calibration()
print(device_calib.get_device_subtype())

Accessing 6DoF poses of sensors with Sophus Python binding​

All 6DoF poses (a.k.a. extrinsic parameters) are represented as relative to the device frame. The device frame is a specific sensor frame, identified by the sensor's label. Aria device frame is by default camera-slam-left. We also provide the pose of the central-pupil-frame in the device frame or as relative to a sensor frame.

label = "camera-slam-right"
transform_device_sensor = device_calib.get_transform_device_sensor(label)
transform_device_cpf = device_calib.get_transform_device_cpf()
transform_cpf_sensor = device_calib.get_transform_cpf_sensor(label)

Accessing sensor calibration​

Each sensor on the device may have a corresponding stream in the vrs and may have a corresponding calibration. However, some types of sensors may not have calibration defined for them (e.g. GPS, WPS, bluetooth), and some sensors may not record stream in a specific vrs. For sensor streams where calibration is available, they can be accessed by labels:

# returns None if vrs does not have a calibration
device_calib = provider.get_device_calibration()
sensor_calib = device_calib.get_sensor_calib(label)

More conveniently, you can just do

stream_id = StreamId("1201-1")
calib = provider.get_sensor_calibration(stream_id)

If you know the calibration type, you can also do

# returns None if the calibration label does not exist
cam_calib = device_calib.get_camera_calib("camera-rgb");
imu_calib = device_calib.get_imu_calib("imu-left");

Accessing ET and Microphone calibration​

Note Aria's ET camera stream and audio are special types:

  • Aria's ET stream switches the stream for left and right ET together, thus its calibration is a pair of CameraCalibration.
  • Aria's Audio stream has 7 channels, thus its calibration is an array of seven microphone calib.
# returns None if the calibration label does not exist
et_calib = device_calib.get_aria_et_camera_calib()
print(et_calib[0].get_label())
mic_calib = device_calib.get_aria_microphone_calib()
print(mic_calib[0].get_label())

Python binding for Sophus library​

Sophus Python PyBind implements Python binding for Sophus that provides access to SO3, SE3, interpolate and iterativeMean features.

This Python binding has been submitted to Sophus and will be officially supported by the Sophus Library GitHub repo soon. Once it is available through the Sophus Library, this section will point to Sophus documentation and code, to avoid duplication.

The user interface is inspired by scipy.spatial.transform.Rotation.

Feature list​

  • SO3
    • Initialize with from_quat(), from_matrix(), exp()
    • Convert to functions: to_quat(), to_matrix(), log()
    • Multiplication with SO3 or 3D points
    • Operator [] for setting/getting items with index or slices
    • Inverse, copy, print, and len
    • Function vectorization
  • SE3
    • Initialize with from_quat_and_translation(), from_matrix(), from_matrix3x4(), exp()
    • Convert to functions to_quat_and_translation(), to_matrix(), to_matrix3x4(), log()
    • Multiplication with SE3 or 3D points
    • Get rotation and translation component with rotation() and translation()
    • Operator [] for setting/getting items with index or slices
    • Function vectorization
    • Inverse, copy, print, and len
    • Interpolate between two SE3
    • Iterative mean of a group of SE3

Import Sophus Python binding​

from projectaria_tools.core.sophus import SO3, SE3, interpolate, iterativeMean

Example code​

Example code is provided in projectaria_tools/core/examples/sophus_quickstart_tutorial.ipynb

python3 -m jupyter ensures that the Jupyter comes from the virtual environment that contains the projectaria_tools module.

Vectorization detail​

In Python, we chose to export our Sophus::SO3 as a vector of SO3 objects by binding the cpp object SO3Group defined below. This is because numerical code in Python tends to work with array of values so that the program can be efficient. This approach is inspired by scipy.spatial.transform.Rotation.

Passing a single SO3/SE3 object to C++ code in Python binding​

To allow other Python binding C++ code to take in a single SO3/SE3 object, we built a caster so that, even if we wrap SO3Group/SE3Group in Python, those can be implicitly converted to the C++ Sophus::SO3/SE3 object at the boundaries between languages.

This enables us to pass the Python SO3/SE3 object to a C++ function as if they were a regular 1-element Sophus::SO3/SE3 object. This simplifies binding the rest of the C++ code. The implicit cast fails if the Python object is not a 1-element object.