DrawerDissect: Whole-drawer insect imaging, segmentation, and transcription using AI

Elizabeth G Postema, Leah Briscoe, Chloe Harder, George R. A. Hancock, Lucy D. Guarnieri, Tony Eisel, Kelton Welch, Nicole Fischer, Christine Johnson, Diego Souza, Dexter Phillip, Rebekah Baquiran, Tatiana Sepulveda, Bruno A S de Medeiros 

Natural history collections curate hundreds of millions of pinned insects worldwide. These invaluable records of biodiversity fuel key research areas, from evolutionary biology to conservation. However, translating physical specimens into images and research-ready data can be prohibitively labor-intensive. To unlock the research potential of insect collections, we developed DrawerDissect, an AI-based pipeline that processes photographs of whole insect drawers, automatically generating dorsal images, background-less (masked) specimens, size measurements, and transcribed metadata. DrawerDissect is available via GitHub: github.com/EGPostema/DrawerDissect. We used DrawerDissect to digitize the Field Museum’s (FMNH’s) entire tiger beetle (family Cicindelidae) collection, resulting in 13,484 high-resolution dorsal photographs, masked specimen images, and body measurements. All specimens are linked to automatically transcribed taxonomic and biogeographic data, as well as specimen-level metadata when visible: in total, 17,264 DarwinCore fields across 12 categories. Compared to a subset of hand-transcribed records, we found that DrawerDissect captures higher-level geographic information most accurately, e.g. Country, State/Province, and County. Finally, we demonstrate the immediate research utility of DrawerDissect outputs with three example use cases: (1) an analysis of taxonomic and environmental drivers of color traits among Cicindela formosa subspecies, using images and metadata from over 300 specimens, (2) an identification model that can accurately classify individuals of two morphologically distinct genera (Cicindela and Staphylinidae: Platydracus) to species, and (3) a country-level inventory of weevils in the subfamily Apioninae using cellphone photographs.