Global Wheat Head Detection Challenge

Intro

Wheat is one of the most important staple crops of the world. High-throughput field phenotyping along with plant breeding helps identifying new cultivars that are resilient to abiotic and biotic pests and at the same time provide sufficient yield. One important yield component are number of wheat heads per m2. But counting wheat heads in high-throughput is a tedious task. Thus, there is a global need for robust and accurate automated counting of wheat heads from images.
Therefore, a competition was launched by several research organizations to find the best algorithm that can do this for a variety of different cultivars from around the world. The crop science group of ETH Zürich provided several hundred pictures with more than 50’000 annotated wheat heads for this competition. Overall, the dataset contains approximately 190’000 wheat heads on 4’700 high-resolution RGB images.

So far (July 2020), more than 1800 teams have participated in the challange.

Aims

• Data Scientists, hackers, and all curious citizen scientists are invited to join forces with us to solve this challenge. An international consortium, has made pictures of more than 190’000 wheat ears available for this competition.
• The competition will run on the Kaggle platform from April to August 2020. Participants are invited to submit software models, based on this dataset, for counting wheat ears effectively.
• The first prize is US$ 15’000, sponsored by the Global Institute for Food Security at the University of Saskatchewan, Canada.
• For full details on the competition and on how to participate, visit: external pagewww.kaggle.comcall_made and external pagewww.global-wheat.comcall_made

Partners

• CAPTE (INRAe - Arvalis - HIPHEN, external pagehttp://umt-capte.frcall_made)
• The University of Tokyo, NARO (Japan, external pagehttps://www.u-tokyo.ac.jpcall_made)
• The University of Queensland (Australia, external pagehttps://agriculture.uq.edu.au/call_made)
• The University of Saskatchewan (Canada, external pagehttps://www.cs.usask.cacall_made),
• Rothamsted Research (Great Britain, external pagehttps://www.rothamsted.ac.uk/call_made)
• ETH Zürich (https://kp.ethz.ch/GlobalWheatDataset/)

Funding and Support

• Provided by partners

Output

external pageGlobal Wheat Head Detection (GWHD) dataset: a large and diverse dataset of high resolution RGB labelled images to develop and benchmark wheat head detection methodscall_made