New Mexico Aerial Waterfowl Surveys: Image Boundaries and Bird Detections

We conducted aerial imaging surveys of waterfowl at wildlife management areas in the Middle Rio Grande Valley of New Mexico between 2018 and 2023 using two unoccupied aerial systems: a DJI Mavic Pro 2 (rotary wing) and a Wingtra One Gen II (fixed wing). In this dataset are projected image boundaries generated in Agisoft Metashape Pro and projected points of bird locations derived from a custom YOLOv5 convolutional neural network model and spatially projected using Geograypher (https://github.com/open-forest-observatory/geograypher/tree/main). The multiclass detector classified birds as duck, goose, or crane. Redundant locations of the same animal in spatially overlapping frames are preserved. These products were used to assess platform bias and count error.