Selective focus shot of a bumblebee sitting on a lavender

bumblebee, flower, purple, insect, nature, bee, summer, macro, plant, garden, pollen, green, closeup, blossom, pollination, flora, spring, nectar, bloom, yellow, animal, natural, background, floral, bumble, wildlife, beautiful, violet, beauty, color, wild, season, flowers, wing, blue, lavender, outdoors, bug, close, petal, field, botany, blooming, small, meadow, wallpaper

Helmholtz AI project call showcase: Pollination Artificial Intelligence

Transferring existing large-scale AI methods to the field of pollination ecology for the automatic classification of insect pollinators.

How can AI methods be used in the field of pollination ecology and taxonomy? Learn more in today’s Helmholtz AI project showcase ‘PAI - Pollination Artificial Intelligence’, a collaboration between the German Aerospace Center (DLR) and Helmholtz Centre for Environmental Research (UFZ).

Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?

Hi, my name is Tiffany Knight and I work on Pollination ecology at the Helmholtz Centre for Environmental Research - UFZ. Our project is titled PAI - pollination artificial intelligence.

In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?

PAI brings together experts in artificial intelligence (AI) and machine learning (ML) with those in pollination ecology and taxonomy to develop automated classification of insect pollinators. This research will enable rapid and consistent monitoring of pollination and the ecosystem services it provides. Pollination is currently monitored using field observations of insects landing on flowers combined with time-consuming identification of these insects by taxonomists. In order to develop automated pollinator identification tools, it is critical that pollination ecologists define the classification groups. Many pollinators cannot be identified to species from images, because distinguishing traits are not visible from images. Thus, higher levels of classification (i.e., genus- and family-levels) are necessary. Our machine learning approach aims to classify expert-defined groups.

The risk is due to the rather challenging classification task. There are thousands of species of insects involved in pollination in Europe. Images from the field have noisy backgrounds and there are often multiple insects on a single flower at once. Insects of the same species can look very different across locations, seasons and genders. Individuals of different species can look very similar. Our project has the potential for high gain, as successful AI tools have the potential to transform the field of pollination ecology.

How important has the Helmholtz AI funding and platform been to carry out this project?

The Helmholtz AI funding program stimulated pollination ecologists to look across Helmholtz Centres for researchers with expertise in machine learning and AI. We found the perfect partners at the DLR. One of the scientists on the team describes himself as someone that loves to solve complicated machine learning problems.

Figure: Image of a bumblebee (black square) pollinating a flower. For the given image, PAI will provide the following results: Insect visitor detected: yes, Order: Hymenoptera, Family: Apidae, Genus: Bombus