Was a bit dissapointed. Thought he was going to do an audio sweep and use ML on a video feed to detect the resonant frequency and break the glass as fast as possible.
I agree it would have been cooler if he did an autodetection sweep (or the impulse response thing mentioned upthreads), but not disappointingly so. However I would have been disappointed if he used ML for the detection. Some basic DSP would suffice and be much more accurate without needing a days worth of training on a powerful video card.
