About BeeMachine

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 with data primarily from the Global BIodiversity Infrastructure Facility (GBIF). We are also grateful for images donated 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 that share their images and taxonomic expertise.

Our Computer Vision Model
The current version of BeeMachine is able to identify bees from around the world, although it works best for bumble bees. Overall test accuracy is 93.4% (99.2% top-3) but this varies by species depending on the number of training images and their level of morphological variability (see fig below). BeeMachine uses a convolutional neural network, modified from EfficientNetV2, and was trained on over 900,000 bee images.

The 108 Bombus species currently recognized by BeeMachine: 

Bombus affinis
Bombus alpinus
Bombus appositus       
Bombus ardens
Bombus argillaceus 
Bombus auricomus
Bombus balteatus
Bombus barbutellus 
Bombus beaticola
Bombus bellicosus 
Bombus bicoloratus 
Bombus bifarius 
Bombus bimaculatus 
Bombus bohemicus 
Bombus borealis 
Bombus californicus 
Bombus caliginosus 
Bombus campestris 
Bombus centralis 
Bombus citrinus 
Bombus confusus 
Bombus consobrinus
Bombus crotchii
Bombus cryptarum
Bombus dahlbomii
Bombus deuteronymus
Bombus distinguendus    

Bombus diversus
Bombus ephippiatus
Bombus eximius
Bombus fervidus
Bombus flavidus
Bombus flavifrons
Bombus formosellus
Bombus fraternus
Bombus frigidus
Bombus funebris
Bombus griseocollis
Bombus haematurus
Bombus honshuensis
Bombus hortorum
Bombus hortulanus
Bombus humilis 
Bombus huntii
Bombus hypnorum 
Bombus hypocrita 
Bombus ignitus 
Bombus impatiens 
Bombus insularis 
Bombus jonellus
Bombus kirbiellus 
Bombus lapidarius 
Bombus lapponicus
Bombus lucorum   

Bombus magnus
Bombus mckayi
Bombus medius 
Bombus melanopygus 
Bombus mesomelas 
Bombus mexicanus 
Bombus mixtus 
Bombus monticola 
Bombus morio
Bombus morrisoni 
Bombus muscorum 
Bombus nevadensis 
Bombus norvegicus
Bombus occidentalis 
Bombus opifex 
Bombus pascuorum
Bombus patagiatus 
Bombus pauloensis 
Bombus pensylvanicus 
Bombus perplexus 
Bombus pratorum 
Bombus pullatus 
Bombus pyrosoma 
Bombus quadricolor 
Bombus robustus 
Bombus rubicundus 
Bombus ruderarius   

Bombus ruderatus
Bombus rufocinctus 
Bombus rupestris
Bombus sandersoni 
Bombus schrencki 
Bombus semenoviellus 
Bombus sichelii
Bombus sitkensis 
Bombus sonani
Bombus soroeensis 
Bombus sporadicus 
Bombus subterraneus 
Bombus sylvarum 
Bombus sylvestris 
Bombus sylvicola 
Bombus ternarius 
Bombus terrestris 
Bombus terricola
Bombus transversalis
Bombus vagans
Bombus vancouverensis 
Bombus vandykei 
Bombus vestalis 
Bombus veteranus 
Bombus vosnesenskii 
Bombus wilmattae 
Bombus wurflenii      

Other bees recognized by BeeMachine: 

Agapostemon
Amegilla
Andrena
Andrena cineraria
Andrena clarkella
Andrena denticulata
Andrena flavipes
Andrena fulva
Andrena haemorrhoa
Andrena hattorfiana
Andrena vaga
Anthidiellum
Anthidium
Anthidium manicatum
Anthophora
Apis
Apis mellifera
Augochlora
Augochlorella
Augochloropsis

Brachymelecta
Braunsapis
Calliopsis
Centris
Ceratina
Chelostoma
Coelioxys
Colletes
Ctenonomia
Dasypoda
Diadasia
Dianthidium
Dieunomia
Duforea
Epeoloides
Epeolus
Eucera
Euglossa
Eulaema
Habropoda

Halictus
Halictus ligatus
Halictus scabiosae
Heriades
Hoplitis
Hylaeus
Lasioglossum
Macropis
Megachile
Megachile sculpturalis  Megachile xylocopoides
Melecta
Meliponini
Melissodes
Melitta
Nomada
Nomia
Osmia
Osmia bicornis
Panurgus

Paranthidium
Peponapsis
Perdita
Plebeia
Ptilothrix
Rhodanthidium
Sphecodes
Stelis
Svastra
Thyreus
Trachusa
Triepeolus
Xylocopa
Xylocopa micans
Xylocopa sonorina
Xylocopa tabaniformis
Xylocopa violacea
Xylocopa virginica      

Learned features

Neural networks learn their own set of features in images to differentiate species. Each point in this visualization represents a test image. Images with similar learned features are close together, forming clusters that correspond well with species. Thus learned features parallel real life qualities that we use to differentiate species.