UC Santa Barbara announced on Mar. 26 that it is leading a collaboration of 13 U.S. institutions in the Big Bee Project, an initiative aimed at digitizing and analyzing bee specimens using advanced technologies such as artificial intelligence, big data, and networked databases.
The project seeks to modernize how scientists study natural history collections by creating over one million high-resolution images and annotated datasets of bee traits. The National Science Foundation provided $3 million in funding for three years to support this work.
Katja Seltmann, director of UC Santa Barbara’s Cheadle Center for Biodiversity & Ecological Restoration, said, “How can we apply these techniques to natural history collections, especially when much of the intrinsic information a specimen has to offer is difficult to quantify?” Her team used machine learning and computer vision alongside crowdsourcing efforts. For example, more than 5,000 volunteers contributed body measurements from detailed bee photos uploaded to the Notes From Nature database. Seltmann noted that volunteer input was comparable to trained scientists.
In another part of the project, researchers used computer vision tools to quantify hair density and color across hundreds of bee species. Their findings were published in a paper detailing correlations between hair coverage and bees’ adaptation strategies for different climates. Seltmann also collaborated with engineering professor B.S. Manjunath on automating wing venation analysis—a trait useful for species identification but traditionally requiring expert review.
UC Santa Barbara’s role has led to dozens of publications and new standards for researching museum collections. “So scientists have really been pushing to turn these into quantitative things, like numbers, matrices and graphs,” Seltmann explained regarding efforts to make complex biological traits measurable with modern data science tools.
Looking ahead, while the current phase of the Big Bee Project is ending, Seltmann plans continued work expanding access to their dataset across disciplines such as engineering and material science.
