Watch the video "12:16 p.m. • Around D.C." and you can clearly see content-aware deep fake filling of crowds to increase the appearance of their magnitude. The so-called
"Parler dataset" is completely useless as anything but a curiosity with respect to examining the nature of emerging digital ground-truth and perception metrics. Nothing contained in it is verifiable, and in my opinion it would be wise to be very skeptical of anything you encounter regarding it.