SDK Reconstruction::addFrame() T parameter

I’d like to know more about the T parameter in the Reconstruction::addFrame() function. Does the “camera coordinate system” mentioned in the docs relate to the main camera or the camera for which the the addFrame() function is being called?

My goal is to improve the reconstruction of a rotating subject by using a known turntable position to improve tracking.

To get an idea of how using a custom T works, my first experiments take the trackedT returned from adding a frame to the first sensor, and using that for the T when adding a frame to other sensors. I am getting very messy results with this. Any reasons why?

The camera coordinate system is always the one of the reference sensor. The transformation from the current to the reference sensor is passed in via the sensorT parameter. So if you have two static cameras you would use the same T parameter for both of them, but for the reference camera sensorT would be nullptr (or identity) and for the non-reference it would be the sensorT you calibrated beforehand.

1 Like