In this work led by Josh Wilson, we labelled 4824 drone-based images containing 49990 birds from 101 species and trained a bird detection and ID model. The model achieved precision = 0.91 for detection & 0.65 for species/age/sex classification.
zslpublications.onlinelibrary.wiley.com/doi/10.1002/...
Drones are a valuable tool for surveying birds, but manually detecting and identifying birds in drone images is costly. We assembled a diverse dataset of 23 865 images of birds captured with 21 diffe...
