main (left) camera transform

How to find main (left) camera transform from world anchor? (Enterprise API) From CameraFrameProvider() I can get a frame sample which has an "extrinsics" parameter. How is it defined? Relative to what point/anchor?

Answered by Vision Pro Engineer in 812836022

The camera extrinsic transform is relative to the DeviceAnchor.

Accepted Answer

The camera extrinsic transform is relative to the DeviceAnchor.

Does it make sense that the extrinsic translation is just 2 centimetres from the device anchor?

translation = 0.024845406, -0.02110077, -0.057464134

From measuring the camera location on VisionPro, it's definitely more than 2 cm from the "center" of VisionPro

From measuring the camera location on VisionPro, it's definitely more than 2 cm from the "center" of VisionPro

The DeviceAnchor origin is not at the "center" of Vision Pro.

  1. what is the DeviceAnchor? Is it the location from which virtual content (and camera passthrough) is rendered on the user's display(s)?

  2. My understanding of a 4 by 4 transformation matrix is that the last column represents the translation, while the first 3 columns represent the local x, y, and z axis. I have plotted the first three columns of the extrinsics matrix and they do not seem to be orthogonal to each other, what are the implications of that? I would expect just a translation and a rotation (without scaling/shearing)

  3. what is the coordinate system for these intrinsics? based on the first 3 columns, it seems to be x=right, y=down and z=forward. Is that correct?

any help or link to well-written documentation very much appreciated. When I see the following documentation which only tells me the type of extrinsics, and literally nothing else than just the type, I find that insufficient. https://developer.apple.com/documentation/arkit/cameraframe/sample/parameters/4443449-extrinsics

my apologies, they ARE orthogonal to each other, I made a sign mistake in my plotting. Still, I'd be curious to know the coordinate system. Thank you

main (left) camera transform
 
 
Q