Assistant Professor Kansas State University, Kansas, United States
Background/Question/Methods
Reliable identification of pollinators, such as bees, is critical for basic ecological research, conservation, and maintaining pollination services. However, species-level bee identification is difficult and requires specialized taxonomic knowledge because of their great diversity and often subtle morphological differences between species. The identification process results in a bottleneck that is expensive and time consuming, which slows the pace of research. However, rapidly developing technology in the fields of machine learning and computer vision are enabling fast and reliable detection and classification of objects, such as bees, from images. We used convolutional neural networks, including InceptionV3, to develop a classification algorithm that can identify bumble bee species from images. Overall species-level accuracy was greater than 92%.
Results/Conclusions
The classifier was incorporated into a mobile app, BeeMachine, that can be used by researchers and community scientists and is now available for Android and iOS. In the future, we hope to use data from contributed sightings to build databases for conservation research. We also hope to incorporate our classification models into automated monitoring systems for large scale monitoring efforts.
