Drones for Ducks

- Project -

US Fish and Wildlife Service
  • Christopher  Lippitt
  • Grant  Harris
  • Rowan  Converse
  • Jesse  Sprague
  • Mohammad Mustafa  Sa'doun
Ducks and geese in water at the Bosque del Apache National Wildlife Refuge surrounded by computer-drawn polygons


We are collaborating with the US Fish and Wildlife Service (USFWS) to develop a more efficient and less invasive method to count and identify species of migratory waterfowl at wildlife refuges across New Mexico. Imaging surveys were conducted using unmanned aerial vehicles (UAVs) at Bosque del Apache and Maxwell National Wildlife Refuges to census migratory birds. Birds were then manually counted and identified by both professional biologists and citizen scientists using a platform called LabelBox. At the present we are investigating 1) whether training data from a smaller sample of professionals or a larger sample of citizen scientists results in more robust training data; 2) how to build a count-based algorithm that can be efficiently re-used with new data year after year; and 3) how to build an identification algorithm with precision comparable to or greater than human observers. 

If you'd like to participate in labeling and identifying waterfowl from imagery, check out our project on Zooniverse! No birding knowledge or experience required!