Well done guys! Really cool to see it launched.
I'm pretty sure you've thought of this but if not here is some algorithm advice: you should start recording baseline density referenced by time (maybe use Fourier to detect time patterns?) so you can start filtering out the noise. That way you can filter out the few thousand pictures of the Empire State Building everyday unless something happens at the ESB in which case density would dramatically increase relative to time.
Also, this would be great to see visually on a map where when you zoom farther and farther in the clusters get redrawn so I can see micro events happening within the major event.
There are so many interesting and potential uses from such a product.
Maybe it would help filter out less relevant results if events without any photos featuring large groups or crowds of people are removed, since those typically go hand-in-hand with notable events. If there could be a way to separate out sporting and music events, that would be nice as well.
Prime featured content would be large groups of people in the streets, with bonus points if there are things on fire. Another would be natural disasters (floods, tornadoes, hurricanes, earthquakes, tsunamis), which most likely do not contain pictures of crowds in the streets, but probably do contain damaged or flooded buildings. Another might be a disease outbreak, perhaps with pictures of individuals with lesions or other visible symptoms, maybe in a medical setting. War zone events like the shelling of a neighborhood somewhere would be interesting, and might fall under the damaged building filter for natural disasters. Maybe pictures of tanks or other heavy military equipment rolling into a town, or helicopters flying overhead. UFO sightings. Pictures of crashes and accidents are always popular as well, and might only have a few emergency personnel around. Political events would probably fall under the large crowd filter. What else?
Congrats guys! ;)