Object Detect Questions

I’m experimenting with Face Tracking using the DetectObject freeframe node.

I have it working pretty well - thanks guys!

A few noobie questions though:

I want to send the results (coordinates) to another software via osc. If there are two faces detected how do I differentiate between the results for each? Is it something to do with spreads?

Is it possible to combine the frontal & profile xml training files so as I can detect a broader selection of faces? Or would I have to have two instances of the detectobject node.

Has anyone got any tips for optimizing tracking reliabilty
eg Lighting/Background advice - bright/daylight/white/blue bground?

Can anyone suggest an effective back up tracking method for if the detect object fails? eg use colortracker if no faces are detected in a frame.

Thanks for any ideas,

Gav