%0 Figure %A Yu-Min Lin, Albert %A Huynh, Andrew %A Lanckriet, Gert %A Barrington, Luke %D 2014 %T Data Collection Results: %U https://plos.figshare.com/articles/figure/_Data_Collection_Results_/1281735 %R 10.1371/journal.pone.0114046.g004 %2 https://plos.figshare.com/ndownloader/files/1857556 %K training mechanism %K consensus %K scale survey %K Genghis Khan %K 10 K %K kernel density estimation %K networked groups %K anomaly detection %K feature categorizations %K landscape %K peer feedback loop %K Genghis Khan Massively %K participant %K reasoning engine %K knowledge generation %K National Geographic expedition %X

Hundreds of thousands of tags overlaid on a subsection of the search area to generate maps of roads (red) and rivers (blue), and to locate ancient (yellow), modern (grey) and other (green) structures. Locations of high agreement (global KDE) in the “ancient structures” category are represented with increased radius. Five example positive identifications are highlighted with labels Fig. 5a- 5e. To construct this figure, we computed the global KDE and applied our visualization function on density peaks across the region. Satellite imagery provided courtesy of the GeoEye Foundation.

%I PLOS ONE