• Hardware
  • Orientation of IMU in different OAK models

I found old posts related to the IMU orientation, but not specifically what I'm looking for.

I used different models of OAK-D cameras, specifically old USB and PoE OAK-D (gen 1), and I used the raw accelerometer on the IMU to find pitch and roll angles after converting the axis orientation to the coordinate reference system that I use.

I recently purchased an OAK-D lite and noticed that the axes on the IMU seem to be orientated differently, so that to obtain the results I need to convert them in a different way to what I used before.

These are the orientations from what I could determine by looking at the raw accelerometer values (excuse my crude drawing, the box on the left is supposed to represent a generic OAK-D camera facing in the viewer direction)

Is this a difference between all gen1 and gen2 models with IMUs? Is there a way to know how the IMU is oriented when connecting to the camera?

    • Best Answerset by erik

    Hi @Letty ,
    On newer devices (manufactured in last 6 months), we also flash imu extrinsics to the device, and you can query them via:

        calibData = device.readCalibration()
        print(calibData.getImuToCameraExtrinsics(dai.CameraBoardSocket.CAM_A))

    ATM it doesn't seem it auto-populates (from device name/revision) these extrisnics if they aren't flashed in calibration, but I believe this is WIP. Let me ask the team. You can find them in depthai-boards repo, example:
    luxonis/depthai-boardsblob/main/boards/OAK-D-LITE.json#L53
    Thanks, Erik

Hi @Letty ,
On newer devices (manufactured in last 6 months), we also flash imu extrinsics to the device, and you can query them via:

    calibData = device.readCalibration()
    print(calibData.getImuToCameraExtrinsics(dai.CameraBoardSocket.CAM_A))

ATM it doesn't seem it auto-populates (from device name/revision) these extrisnics if they aren't flashed in calibration, but I believe this is WIP. Let me ask the team. You can find them in depthai-boards repo, example:
luxonis/depthai-boardsblob/main/boards/OAK-D-LITE.json#L53
Thanks, Erik