> Then it sounds like they used a DNN or something similar to analyze the rate of fireball expansion in the films to refine their earlier estimates of yield.
It's actually quite amazing that even with such a simple model, G.I. Taylor was able to estimate yield's to within 10% of the official figure for the Trinity device.
Just because DNN's are terribly trendy doesn't mean that every image analysis task is done with them. :) I think that at least with such a crude model as the Taylor one, the model itself will introduce errors that DNN's per se won't fix. DNN's might help the manual drudgery of analyzing each frame of the videos manually though..
Back in the day, some rather simple models were used to estimate yield based on photographs, see https://en.wikipedia.org/wiki/Nuclear_weapon_yield#Calculati...
It's actually quite amazing that even with such a simple model, G.I. Taylor was able to estimate yield's to within 10% of the official figure for the Trinity device.
Just because DNN's are terribly trendy doesn't mean that every image analysis task is done with them. :) I think that at least with such a crude model as the Taylor one, the model itself will introduce errors that DNN's per se won't fix. DNN's might help the manual drudgery of analyzing each frame of the videos manually though..