>BeeMachine was developed by Brian Spiesman in collaboration with:

Claudio Gratton, University of Wisconsin – Madison
William Hsu, Kansas State University
Brian McCornack, Kansas State University

>Support 
BeeMachine was funded by USDA NIFA and Kansas State University. Computer vision models were developed using Global Biodiversity Infrastructure Facility (GBIF) data. We are also grateful for data provided by others, including the Wisconsin Bumble Bee Brigade, the Hanamaru Maruhana Project, and Jerry Cole. I am grateful for the volunteer participants in these programs who shared their images and taxonomic expertise. 

>Our Computer Vision Model 
BeeMachine can identify bee species from around the world. But it will give a genus-level prediction if it isn't sure about the species. Because flowers are often visited by other insects that can sometimes be confused with bees, we now include the ability to differentiate bees from wasps, flies, beetles, and butterflies/moths. Overall test accuracy on the current algorithm is 93.7% (99.4% top-3), but this varies by species depending on the number of training images and their level of morphological variability (see figures below). BeeMachine uses a convolutional neural network, modified from EfficientNetV2, and was trained on over 1.2 million images. Neural networks learn their own set of features in images to differentiate species. 


I loaded the BeeMachine app, and it works great.  It occurred to me that with Merlin from Cornell for bird calls, iNaturalist for plants, and BeeMachine, I now have instant access to resources that would have taken hours to research.   

             ***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software.  For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html