Community & Landscape Ecology: BeeMachine v1.0

An important part of bee conservation for gardeners and growers alike is the construction of bee-friendly spaces. A large part of creating a successful pollinator garden is to choose the right plants that will promote visitation. To do this, researchers have to figure out which plants are “attractive” to bees, and which species of pollinators are attracted to which plants. Unfortunately, in an outdoor setting, this can be extremely difficult for the layperson and entomologist alike.

Brian Spiesman of Kansas State University has developed a program called the BeeMachine. In his words the BeeMachine implements “leading edge computer vision such as convolutional neural networks to detect and classify bees captured in images and video. We are working on developing models for automated species-level identification and deploying sensors for automated sampling and observation of plant-pollinator interactions”. As of this writing, 36 species of bumble bees can be identified with accuracy using this program. Simply put, BeeMachine allows someone to take a picture or video of a bumble bee and upload it to the BeeMachine website. Following the upload, the website will output three species that have the highest probability of being the bumble bee in question. An example of the output can be seen here:

Bumble bees can be found throughout North America and are extremely important pollinators whose ecosystem services must be protected. BeeMachine can potentially be an invaluable resource when constructing and maintaining a pollinator-friendly space.

The BeeMachine website can be found here:

Brian Spiesman’s research interests can be found here: