Data-centric AI approach for automated wildflower monitoring

Both researchers and policy makers are in need of standards and tools that help understanding and assessing natural capital. Wildflowers are a major component of our natural capital; they play an essential role in ecosystems, improve soil health, supply food and medicines, and curb climate change. In this paper, we present the Eindhoven Wildflower Dataset (EWD) as well as a PyTorch object detection model that is able to identify and count wildflowers. EWD, collected over two entire flowering seasons and expert annotated, contains 2002 top-view images of flowering plants captured ‘in the wild’ in five different landscape types (roadsides, urban green spaces, cropland, weed-rich grassland, marshland). It holds a total of 65571 annotations for 160 species belonging to 31 different families of flowering plants and serves as a reference dataset for automating wildflower monitoring. To ensure consistent annotations, we define specific floral count units (largely based on inflorescences) and provide extensive annotation guidelines. With a 0.82 mAP (@IoU > 0.50) score the presented baseline model, trained with a balanced subset of EWD, is to the best of our knowledge superior in its class. Our approach empowers automated quantification of wildflower richness and abundance and encourages the development of standards for AI-based wildflower monitoring. The annotated EWD dataset is publicly available on the DataverseNL research data repository, and the code to train and run the baseline model is supplied as supplementary material.