The Center for the Advancement of Spatial Informatics Research and Education (ASPIRE) at the University of New Mexico is collaborating with US Fish and Wildlife to develop an algorithm to automatically count and identify birds from drone imagery. We're hoping to create a more efficient + accurate process for counting populations of migratory birds than the current method of counting birds from planes + ground. Humans both in the air and on the ground tend to undercount birds by up to 30% when these counts are compared with counts of birds in imagery. However, counting birds in imagery is extremely time consuming, so we want to automate the process!
To start with, we're comparing sets of images from wildlife refuges in New Mexico where birds have been identified by US Fish and Wildlife biologists to sets of images with identifications by citizen scientists to train the bird counting algorithm how to separate: 1) birds from not-birds (rocks, trees, etc) 2) different species of birds from one another (eg, ducks from geese).
No expert birding knowledge required, if you can tell a duck from a crane, you can try your hand at this project!
