I’m having trouble getting a reading from my bounding box confidence. It’s getting zero input, but there’s input/info coming in from all other individual channels.
Could you give a screengrab of your CHOP and its parameters? Do you mean the confidence channels have a value of 0?
Unfortunately I think this is a limit of the Nvidia AI model that is being used with those settings. We are requesting the bounding box confidence from the model, but all we get are zeroes.
The Face Track CHOP uses different models depending on what parameters are turned on. In the default mode, it uses the FaceDetection model and the bounding box confidence seems to work in that mode.
However, if you enable rotations, landmarks or the mesh transformations than it switches to different models and it seems like those do not return valid bounding box confidence numbers.
Maybe you could turn on landmark confidence values and use ones of those instead?