BirdCollect: A Comprehensive Benchmark for Analyzing Dense Bird Flock Attributes

BirdCollect introduces the largest annotated bird dataset, revolutionizing wildlife research with over 6,000 high-resolution images of dense bird flocks, including Demoiselle Cranes and 34 other species from Keoladeo National Park. This benchmark dataset, spanning 46.2 GB, provides detailed annotations for tasks like crowd counting, density estimation, segmentation, and species classification, addressing the challenges posed by varied conditions such as lighting and resolution. The dataset highlights the need for tailored AI approaches in wildlife monitoring, offering unprecedented insights into bird behavior and diversity along migratory flyways.

., K., Shreshtha, S., Dutta, B., Dosi, M., Vatsa, M., Singh, R., Anand, S., Sarkar, S., & Mali Parihar, S. (2024). BirdCollect: A Comprehensive Benchmark for Analyzing Dense Bird Flock Attributes. Proceedings of the AAAI Conference on Artificial Intelligence, 38(20), 21879–21887. https://doi.org/10.1609/aaai.v38i20.30189