What is the FOV in the factory calibration

chintal

I'm using an OAK-D Lite.

The hardware spec of the OAK-D cameras specify three FOVs - HFOV, VFOV and DFOV.
I see from the docs that readCalibration -> getFov() only takes a camera socket and returns a float.
I also see from this example that the returned FOV seems to be the HFOV.

Where in the documentation does it say this is the HFOV? I've tried looking and can't seem to find it.
Does this example use the number for HFOV to calculate VFOV by assuming the pixels are square and extrapolating from HFOV by frame shape ? Is this some other FOV metric entirely?
Is there a way to get VFOV and DFOV from the camera itself?
How 'reliable' are the factory calibrations of these cameras? Is it expected that the user will redo the calibration, or can I rely on the provided data from the device EEPROM to get somewhat respectable answers?

jakaskerl

Hi chintal

chintal Where in the documentation does it say this is the HFOV? I've tried looking and can't seem to find it.

It doesn't, but FOV generally means HFOV.

chintal Does this example use the number for HFOV to calculate VFOV by assuming the pixels are square and extrapolating from HFOV by frame shape ? Is this some other FOV metric entirely?

That's one way, the other is calculating it from intrinsics.

chintal Is there a way to get VFOV and DFOV from the camera itself?

Usually having the HFOV, you can calculate VFOV from aspect ratio and DFOV from diagonal of those two.

chintal How 'reliable' are the factory calibrations of these cameras? Is it expected that the user will redo the calibration, or can I rely on the provided data from the device EEPROM to get somewhat respectable answers?

As in extrinsics and intrinsics? Very. As in specFov? Not so much. I suggest when retrieving the FOV, you use useSpec=False so FOV is manually calculated from intrinsics.

Thanks,
Jaka