Credits: Global Wheat Head Detection Challenge

作为小麦产量“三因素”之一，单位面积穗数的调查举足轻重，然而，这一重要的产量性状却仍然是人工调查为主，费时费力。近些年，应用图像分析以及深度学习等技术，田间穗数的自动计算已经有不少研究提出，但是由于各个研究都只是聚焦于本地实验数据，数据量小，数据多样性不足，导致至今没有一个通用模型。

于是，经过7个国家、9所科研院所、10多位表型专家长达一年的精心准备，先进级小麦穗识别图像数据库Global Wheat HEAd deTection (Global WHEAT)诞生啦！利用这个数据库，我们计划于2020年5月4日在KAGGLE上举行首届Global Wheat Head Detection Challenge！ 
欢迎CV界高手的参加！

一等奖 8,000 USD奖金总额15,000 USD！（由加拿大Global Institute for Food Security at the University of Saskatchewan赞助）

本次挑战也是表型组织IPPN倡导的CVPPP 2020（COMPUTER VISION PROBLEMS IN PLANT PHENOTYPING）的官方挑战赛，获奖者也会受邀投稿，参加计算机视觉领域先进会议ECCV2020！

本次挑战赛收到了加拿大GIFS，日本Kubota，法国DigitAG，Hiphen，日本Quantomics，南京农业大学Science合办期刊Plant Phenomics 的资金支持。 CVPPP 2020官方网站

https://www.plant-phenotyping.org/CVPPP2020-CfP

Global WHEAT官方网站

http://www.global-wheat.com/

大赛组委会成员：

· CAPTE (INRAe - Arvalis -HIPHEN, http://umt-capte.fr)

· The University of Tokyo, NARO (Japan, https://www.u-tokyo.ac.jp)

· The University of Queensland (Australia, https://agriculture.uq.edu.au/)

· The University of Saskatchewan (Canada, https://www.cs.usask.ca),

· Rothamsted Research (Great Britain, https://www.rothamsted.ac.uk/)

· ETH Zürich (Switzerland, https://usys.ethz.ch/)
Global WHEAT数据责任人：Wei Guo, University of TokyoEtienne David, Arvalis / INRAESimon Madec, Arvalis / INRAEIan Stavness, University of Saskatchewan

Credits: A – Arvalis, B- INRAE, C- Rothamsted Research, D- ETH Zürich

CHINESE:图像说明 / 各地开发出了不同的田间图像采集方法，有低成本的方法（A）、机器人（B）、龙门架（C）或悬挂式（D）。

ENGLISH: Photo caption / Different methods of open-field image acquisition have been developed throughout the world, with low cost methods (A), robots (B), gantries (C), or cable-suspended(D).

FRENCH: Légende photo / Différentes méthodes d’acquisitions d’images en plein champs ont été mises au point à travers le monde, avec des méthodes low cost (A), des robots (B) , ou bien des portiques (C-D).

JAPANESE: キャプション／様々な圃場における画像の取得方法が中で開発されており、低コストの方法（A）、ロボット（B）、ガントリー（C）、懸垂ケーブル式（D）などがある。

Credit: Etienne David CHINESE:图像说明/ 通过图像分析数麦穗数目，至今仍是一个尚未解决的重大难题。

ENGLISH:Photo caption / Counting wheat head by image analysis remains a major unresolved challenge to date.

FRENCH:Légende photo / Le comptage d’épis par analyse d’images reste un défi majeur non résolu à ce jour.

JAPANESE:キャプション／画像解析による小麦の穂数の計数は、現在でも大きな未解決の課題となっています。

Presentation of Global Wheat Head Detection ChallengeENGLISH

An international computer science competition to count wheat heads more effectively, using image analysis

The ProblemFor several years, agricultural research has been using sensors to observe plants at key moments in their development. However, some important plant traits are still measured manually. One example of this is the manual counting of wheat heads from digital images - a long and tedious job. Factors that make it difficult to manually count wheat heads from digital images include the possibility of overlapping heads, variations in appearance according to maturity and genotype, the presence or absence of barbs, head orientation and even wind.

The Need There is the need for a robust and accurate computer model that is capable of counting wheat ears from digital images. This model will benefit phenotyping research and help producers around the world assess Head density, health and maturity more effectively. Some work has already been done in deep learning, though it has resulted in too little data to have a generic model. The CompetitionThe Global Wheat Head Detection Challenge, an international data science competition, was created to address this need. The objective is to have a software model capable of locating heads on a wide variety of data, without bias. Data Scientists, hackers, scientists and the curious re invited to join forces with us to solve this challenge!

Details:· The competition will run on the Kaggle platform from April to July 2020.· International consortium, Global Wheat Dataset, has made more than 190,000 wheat heads available for this competition. Participants are invited to submit software models, based on this dataset, for counting wheat heads effectively.

Competition organizers· CAPTE (INRAe - Arvalis - HIPHEN, http://umt-capte.fr)· The University of Tokyo, NARO (Japan, https://www.u-tokyo.ac.jp)· The University of Queensland (Australia, https://agriculture.uq.edu.au/)· The University of Saskatchewan (Canada, https://www.cs.usask.ca),· Rothamsted Research (Great Britain, https://www.rothamsted.ac.uk/)· ETH Zürich (Switzerland, https://usys.ethz.ch/)
PrizeThe first prize is 8,000 USD (15,000 USD in total) 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: www.kaggle.com and www.global-wheat.com 

An international competition to better count heads by image analysis

 In the era of digital agriculture, monitoring crops with imaging-based sensors is becoming widely applied in the field. Compared with the conventional manual technique, generally this allows to collect traits that reflect the status of crop growth and development more efficient and with a higher accuracy. Wheat is one of the major staple crops. Among the traits of interest, head density, number of heads per unit area, is crucial to assess the tillering capacity and estimate yield. However, due to the overlapping between heads, it is challenging to investigate head density from digital imagery. Obviously, deep learning technique is suitable to tackle with this problem. This motivates us to initialize this competition, "Global Wheat Head Challenge”. It is hosted by the Kaggle platform, recognized as the first in agriculture domain (website). To achieve a robust model, a dataset of more than 190,000 heads annotated with images from very diverse conditions is made available to participants by the international consortium Global Wheat Dataset (website). The competition is co-organized by UMT CAPTE (INRAe - Arvalis - HIPHEN, http://umt-capte.fr), the University of Tokyo, NARO (Japan), the University of Queensland (Australia), the University of Saskatchewan (Canada), Rothamsted Research (Great Britain), ETH Zurich (Switzerland) and Nanjing Agricultural University (China). Thanks to the support of the Global Institute for Food Security (www.gifs.ca). The competition will be held from 4th May to 4th August. The final result will be released on ECCV conference during CVPPP workshop end of August (to be determined). The top 3 teams will be invited to attend the conference with a prize of $15,000 for the most successful teams. You can find more detailed information on website. We therefore invite Data Scientists, hackers, scientists and the curious to join forces with us to solve this challenge!
About Plant Phenomics

《植物表型组学》（Plant Phenomics）是由南京农业大学和美国科学会（AAAS）合作创办的英文学术期刊，于2019年1月正式上线发行，是Science合作出版的第二本期刊。采用开放获取形式，刊载植物表型组学交叉学科热点领域具有突破性科研进展的原创性研究论文、综述、数据集和观点。具体范围涵盖高通量表型分析的较新技术，基于图像分析和机器学习的表型分析研究，提取表型信息的新算法，作物栽培、植物育种和农业实践中的表型组学新应用，与植物表型相结合的分子生物学、植物生理学、统计学、作物模型和其他组学研究，表型组学相关的植物生物学等。期刊已被DOAJ、CNKI、CABI数据库收录。

编辑：黄艺清 （实习）、孔敏审核：尹欢、陈文珠