Hi, I have a similar issue here.
I’m trying to implement RGB–thermal alignment in DepthAI v3 using an RVC2-based thermal camera and my goal is able to tune staticDepthPlane suing runtime. The thermal stream works completely fine when I use the normal node::Thermal → Sync approach, and I can tune all ambient parameters (distance, emissivity, etc.). However, this Sync-only setup does not allow updating the staticDepthPlane at runtime, since that requires going through ImageAlign.
As soon as I switch to the v3 alignment pipeline:
thermal->temperature.link(align->input);
rgbOut->link(align->inputAlignTo);
align->outputAligned.link(sync->inputs["Raw_Aligned"]);
the aligned thermal output becomes all NaN, and min/max prints as -nan / -nan. No valid thermal pixels are produced. Eventually the device errors out with: Device already closed or disconnected: io error
I also test with readCalibration of thermalsocket, the values are valid so calibration is not the issue.
So my question: Is ImageAlign supported for RVC2 thermal input in v3? Also, is thermal-to-RGB alignment with tunable staticDepthPlane possible in hardware on RVC2, or must it be done entirely on the host?