Per frame camera pose

I tried to export the camera pose for each frame while rendering video capture. I add the code to export each frame’s view and projection matrix of viewport_api in the file capture/viewport/extension.py _capture_viewport function.

My question is what could be the correct fx, fy, cx, cy or the hfov, vfov of the pinhole camera that I set.
I recorded view matrix, projection matrix, transform matrix, aspect_ratios, image width and height.
The image resolution is (1080x1080), but the projection matrix gives
p=[[0.7942293090356872 ,0, 0, 0],
[0., 2., 0., 0.],
[0, 0, -1, -1],
[0, 0, -2, 0]]
which has different value on p[0,0] and p[1,1]. Does it means the pixel is not squared? The aspect ratio has two value [0.3971146545178436, 1] which does not meet the aspect ratio of image resolution.

But using any equation found on the internet, the focal length I calculated is different from what is calibrated by RealityCapture. Can you help to obtain the focal length? I need to use camera intrinsics and extrinsics in my application.

hi @zehao666 , you can obetain the cameraFocalLength by using the cameraParams annotator, which you can find the docs here: Annotators Information — Omniverse Extensions latest documentation