Excited to share this work led by Josh Wilson, a PhD student Tatsuya co-supervised, where we compiled 23,865 drone-based bird images, captured using 21 different cameras across 11 countries, and labelled 4824 of them containing 49,990 birds from 101 species.

To demonstrate the usefulness of this dataset, we trained a bird detection and identification model using the dataset; the model achieved a mean precision of 0.91 for detection and 0.65 for species/age/sex classification.

Congratulations Josh and read the paper here.