We've heard a lot about Artificial Intelligence recently. Or is it Machine Learning? Anyway, humans are evidently getting better at training computers. The World Bank has teamed up with drone and photography NGOs WeRobotics and OpenAerialMap to see if computers can automatically identify tree species (taking jobs, again?). Geeks can compete to make the best code to automatically distinguish trees from overhead photographs. Imagery can then be compared before and after disasters to see how much local agriculture might be affected. The test is tuned for Pacific islands set with aerial photos of Tonga. To win, the AI code should know the difference between coconut, banana, papaya and mango trees 80% of the time, and not mix them up with electricity poles or random objects. Human collaborators have classified 13,000 objects by eye, and the computers have to figure out how to do it themselves. May the best bot win.
- Credits Ben Parker/IRIN/WeRobotics/OpenStreetMap
- Themes Solutions and Innovations
- Regions Tonga