Famous Pentagon mad-scientist bureau DARPA says it would like some miracle software to help hard-pressed US spooks and military intelligence officers trawl through the massive troves of imagery retrieved from gadgetry seized in the ongoing massive campaign of special-forces raids across southwest Asia.
The new project has been dubbed Visual Media Reasoning (VMR) and is the subject of a new solicitation (44-page/364KB PDF) issued this week by the military boffins. In the announcement, DARPA says:
VMR's focus is on extracting as much information as possible from fuzzy, noisy, and ill-posed images that have little or no associated metadata and originate from unknown sources and unknown context. These images may be from captured adversary cameras, laptops, and related devices ...
Another intended use case for the VMR program is longer-term forensic analysis on captured media done by investigators in support of theater operations.
It was widely reported that super-elite US Navy SEALs from the unit variously known as Team Six or the Naval Special Warfare Development Group (aka DevGru) seized a large stash of thumb drives, laptops and other gadgets at Osama bin Laden's Pakistani hideout after shooting the famous terrorist dead. That raid alone probably provided enough data to tie up a large number of human intelligence analysts from the military, CIA etc for a long time.
The elite-of-the-elite "Tier One" operatives of the US Joint Special Operations Command* alone are carrying out more than 1,700 raids in a year at the moment and the many other kinds of special forces active across Afghanistan and sometimes Pakistan, the Horn of Africa etc must be carrying out many more such operations. It seems likely that the haul of gadgetry loaded with data must be quite considerable, hence the DARPA push to automate more of the analysis.
According to DARPA:
Our adversaries frequently use video, still and cellphone cameras to document their training and operations and occasionally post this content to widely available websites. The volume of this visual media is growing rapidly and is quickly outpacing our ability to review, let alone analyze, the contents of every image. The VMR technology will serve as a "force-multiplier" by extracting tactically relevant information for the human analyst and alerting the analyst to scenes that warrant the analyst's expert attention ...
With the VMR solution, a warfighter in one military service branch can use the same cutting- edge facial recognition and geo-location technologies as an analyst in an intelligence agency and can avoid applying multiple software applications to extract different visual information from the same image. Imagine needing to use scores of different internet search engines in order to research a complex subject, because each search engine was limited to a particular type of content ...
If DARPA can crack the problem of automated image analysis, there would of course be many other applications for the tech than merely trawling through Osama bin Laden's personal gadgets. But there have been quite a few such military- and intelligence-funded efforts before, so the debut of VMR is probably more an indication that the need for such kit is growing, rather than a sign that it will soon be here. ®
*JSOC is thought to be made up of just three main operational units: Team 6/DevGru, Delta Force, and the constantly name-changing Intelligence Support Activity. There are tens of thousands of other US "Tier Two" special ops personnel such as ordinary SEALs, Rangers, Green Berets, Marine Force Recon etc – and allied units such as the British SAS, SBS and SFSG to boot.