It sounds a bit odd, but all of this data out there needed for a particular event will be harnessed; however, in this case a vast amount of information has to be called for by whomever, then collected, reviewed manually for relevance and then quality, analyzed manually, and put through facial recognition.
In this situation, time is against law enforcement, because every minute spent with the above process, is a minute wasted finding possible suspects.
If the Department of Homeland Security were to build an archive data warehouse for images and videos , then with assistance from Mobile OS companies to add an option to the existing share button to add images & video to upload to a national law enforcement archive, a relatively trivial patch e.g. for Android and iOS, part of the process battle is there. Also the user would have to opt in to send their media and therefore there are limited privacy issues.
Image and video facial recognition could start immediately and rerun incrementally based on new media uploaded as the trigger. If the user does allow the media to be shared, they can add an event name, or select from a DHS list, put notes if the user chooses, automatically date time stamp the media, optionally GEO stamped, and select anonymous or allow the sending the user’s contact info stored. Now the true human analysis work may begin without many manual steps.