>> Nvidia Jetson Xavier NX
Curious -- since you're only doing the inference/generation on the frame, and since you're not doing it all the time, did you need a Jetson or would an RPI have sufficed? Did you test inference speeds across different edge compute options?