This 2013 challenge was developed in partnership with NASA; University of California, San Diego; and National Geographic. Conducted on Topcoder with a $15,000 prize purse, this machine learning contest asked competitors to develop an algorithm that learns from crowdsourced human annotations of anomalies (i.e. structures) found in satellite imagery.
The task was to study select satellite imagery of the region of the lost tomb of Genghis Khan and then develop an algorithm that will recognize human built, and potentially historically significant, structures found in those images (see the problem statement). The contest lasted three weeks and garnered 39 competitors with 357 submissions. The top result produced a promising re-trainable algorithm that can be taught to recognize various structures from satellite images. Learn more about this challenge on the LISH Innovation Science Guide.