Tamara PattonCrowdsourcing lessons from Tomnod

From Tomnod’s website: Examples of possible military buildings, crowds, landmarks and military vehicles in Damascus detected by Tomnod’s CrowdRank and derived from volunteer imagery insight.

* * *

DigitalGlobe announced today that it is acquiring Tomnod, a five-guy operation known for its pioneering work in the field of crowdsourced intelligence.  By combining satellite imagery and Tomnod’s own unique crowdsourcing algorithms, this group has done some remarkable work in areas as varied as hunting for the tomb of Genghis Khan, to searching for lost hikers in Peru, to tracking the conflict in Damascus.

DigitalGlobe and Tomnod have already worked together in the past, such as Tomnod contributing layers of crowd-derived insight to Digital Globe’s First Look, an online service for emergency management that provides quicker access to imagery of world disasters.  DigitalGlobe’s new acquisition of Tomnod will now result in a more sustained partnership, with Tomnod’s information being increasingly used as a data source for DigitalGlobe’s in-house analytics teams.

Here in the arms control world, the search continues (largely thanks to Rose Gottemoeller) for potential new ways to harness the information beast that is “the crowd” (e.g. tweets, photos, blog posts, media, etc.) and use it to support arm control verification efforts.  Though they haven’t engaged in this area, I think Tomnod has come up with some interesting solutions to challenges similar to those faced when considering possibilities for the role of public verification in arms control.  As some food for thought, here are a few such concerns along with Tomnod’s way of tackling similar problems:

Concern: Incentivizing people to participate in “public verification challenges” would be difficult without financial incentive (which would be too costly to sustain for a long period of time).

Tomnod’s workaround:  Make it fun. With Tomnod’s crowdsourcing endeavor to locate the lost tomb of Genghis Khan in Mongolia, they have managed to attract over 43,000 “online explorers” as volunteers which have in turn processed over 900,000 images.  How did Tomnod amass such a following of free labor?  By making the image tagging process into a learning experience that aims to feel more like a game than work.  So did they find it, you ask?  Not yet apparently, but they have reportedly made a host of other new discoveries.


Concern: Calling on the public to become whistleblowers might inadvertently put private citizens in harm’s way.

Tomnod’s workaround: Create a secure, password-protected environment where users can contribute information. Given the sensitivity of the Syrian crisis, when Tomnod worked to monitor the crisis in Damascus, it employed a “community sourcing” approach with a password-protected site sent to invite-only groups. Volunteers in the community could then sift through the images to tag things like landmarks, a crowd, or military buildings.  This reportedly resulted in a stream of intelligence tips about hotspots in Damascus, delivered in real-time to administrators and analysts via web browsers and mobile devices.


Concern: How to filter out the users who enjoy tweeting false alarms of an incoming North Korean missile attack from those who may have actually sighted something useful.

Tomnod’s workaround: A fancy algorithm.  Tomnod has developed a proprietary algorithm known as CrowdRank™ that helps sift out reliable users and reliable information from the rest of the mess.  Read more about it here.


Developing potential crowdsourcing mechanisms for arms control verification presents some unique challenges that won’t be easily resolved, but as we move forward in our thinking, it’s good to remember that crowdsourcing as a field in itself has been around for a long time and that it is well-developed in other areas and applications. I think there’s a great deal we can learn from companies like Tomnod that have taken crowdsourcing to new levels through integrating the power of the crowd with the advances of satellite imagery and remote sensing.

After all, could there really be that much difference in appearance between a camouflaged warhead storage site and Khan’s hidden grave?  The search goals may differ, but there’s much room for overlap in the mechanisms for achieving them.



  1. John Schilling (History)

    “After all, could there really be that much difference in appearance between a camouflaged warhead storage site and Khan’s hidden grave?”

    The muzzle flashes from the guards’ guns may be a dead giveaway.

  2. Luke (History)

    Great article Tamara. Summarizes the Tomnod vision better than I could — and I’m one of those five guys! I would love to apply Tomnod technology to arms control or any mission where the eyes of the world can leverage satellite imagery to make discoveries that affect our planet.
    Get in touch.

  3. Eve (History)

    Some interesting concepts and it is a good discussion point, but whatever happened to OSGEOINT?

  4. Cthippo (History)

    The frustrating thing about this sort of work is the difficulty in getting anyone to pay attention to the results. It’s been my experience that those who are in a position to use this kind of data are loath to accept it from anyone they consider an amateur.

    My experience with this comes from the realm of hunting for shipwrecks and derelict vessels. I used a number of sources, including cruising around on Google Earth, to find potential wrecks and then take my kayak out to visit and document them. There is a an office at the state which is responsible for removing such vessels and for a long time I sent them photos, GPS coordinates, and data on the wrecks I found, but it eventually became clear that they were completely uninterested in the data I was sending. I suspect they were concerned that eventually I would send them info on a wreck that they actually had to do something about, but who knows.