Gerard Schouten, lecturer in AI & Data, and researcher Bas Michielsen of Fontys University of Applied Sciences have been researching how to use AI to measure the biodiversity of wildflowers in the Netherlands. They collected 2,000 images of wildflowers taken in 5 landscape types (roadsides, urban areas, croplands, grasslands and wetlands) around and in the city of Eindhoven. The pictures were taken with a drone or a modern smartphone. With these images, Schouten and Michielsen trained the AI model.
The images, which contain some 160 flower species, were supplemented with some 65,000 annotations. This annotated dataset is open access available via the Fontys collection in DataverseNL. The dataset can be used to develop AI (computer vision) algorithms that can count and identify wildflowers. It can also be used as a reference dataset to benchmark object detection algorithms. Schouten and Michielsen hope that other researchers will use their dataset to create even better models.
Fontys University of Applied Sciences uses DataverseNL to publish their data. DataverseNL is a research data repository offered jointly by DANS and participating institutions. DANS manages the technical infrastructure and the institutions using DataverseNL are responsible for assigning rights to user accounts, managing and curating the deposited research data within DataverseNL.
An article on the AI model was also published in Bron, a news medium of Fontys Universities of Applied Sciences. The publication on the AI model can be found in PLOS ONE.
More information about DataverseNL can be found here.
