Project Leads: This email address is being protected from spambots. You need JavaScript enabled to view it. , This email address is being protected from spambots. You need JavaScript enabled to view it.
The Global Wheat Head Segmentation project aims to develop practical image-analysis to assess wheat spike morphology in the field. The provided dataset was curated by the Global Wheat dataset community (https://www.global-wheat.com) aiming to provide large, diverse and well-annotated image datasets of wheat.
The GWHS dataset contains ~3000 fully visible wheat heads cropped from field canopies around the world (see image). The peduncle and the terminal spikelet (marking the extremes of the spike) as well as all spikelets and their orientation were pixel-annotated (see picture). The dataset is a good test case for a hackathon, because it requires a multi-step detection and quantification workflow. Fully visible spikes need to be detected within the canopy and their side-branching spikelets segmented and counted. For each spikelet, its orientation needs to be classified, and symptoms of any damage need to be quantified. Such symptoms can be color changes or stunted development and deformation. The GWHS set contains a sizable amount of images showing Fusarium head blight and we will, therefore, focus on this type of damage.
At BioHackathon Germany 2026, we want to move beyond simply training another model. At this hackathon we will focus on designing intelligent, reusable workflows that bring existing computer vision models together into practical tools for plant phenotyping specialists, breeders, and plant pathologists. Thus, participants will not start from scratch. They will work out how existing models and APIs can be combined to assess multiple spike characteristics. The Global Wheat community will provide an annotated GWHS dataset and a collection of existing models through the Global Wheat Hugging Face Space, accessible via Gradio APIs. Preconfigured Docker environments on the Renku platform and a Jupyter Notebook tutorial will help with accessing the models, combining their outputs, exploring the dataset, and getting new workflows up and running.
Smartphone-based phenotyping is where this is heading-giving researchers and breeders a way to extract meaningful spike and spikelet information directly from field images using tools they already carry.
We welcome participants interested in computer vision, foundation models, plant phenotyping, wheat breeding, plant pathology, workflow design, and mobile applications
.