Drones for Ducks: Automated Waterfowl Identification and Counts Portal
This portal is aimed at wildlife managers and researchers interested in deriving automated counts and identification of waterfowl in the southwestern United States from aerial imagery. Get started by creating an account and following the instructions to upload your imagery.
Waterfowl are detected in uploaded imagery using a YOLOV5 machine learning object detection architecture. Three classes based on morphology are currently supported: duck, goose, and crane. The current model was trained using 593 images of waterfowl captured at Bosque del Apache National Wildlife Refuge in 2018 with over 215,000 labels from the citizen science platform Zooniverse. Images that are uploaded to this portal can be labeled automatically within minutes by using the backend trained model!
By default, labeled imagery and an .xlsx spreadsheet with image coordinates and a confidence value for each label will be returned.
Spatial data in the form of an ESRI Shapefile will be returned to show the location of each bird detected.
This portal was produced through Drones for Ducks, a collaboration between the US Fish and Wildlife Service and the Center for the Advancement of Spatial Informatics Research and Education (ASPIRE) at the University of New Mexico. Our goal is to automate the process of censusing waterfowl in the Southwest Unit of the National Wildlife Refuge System through the use of small uninhabited aerial systems (sUAS, aka “drones”) for field surveys and machine learning for automated classification and counting.
ASPIRE works to advance the science of GIS and remote sensing through cutting-edge research and the development of new technologies and methodologies. Learn more about us and our other projects here!
The US Fish and Wildlife Service is a federal agency tasked with the management of wildlife refuges, endangered species, habitat restoration, and other priorities. You can learn more about USFWS and the wildlife refuge program here!