Global Wheat Head Detection 2021: an improved dataset for benchmarking wheat head detection methods

The Global Wheat Head Detection (GWHD) dataset was created in 2020 and has assembled 193,634 labelled wheat heads from 4700 RGB images acquired from various acquisition platforms and 7 countries/institutions. With an associated competition hosted in Kaggle, GWHD-2020 has successfully attracted attention from both the computer vision and agricultural science communities. From this first experience, a few avenues for improvements have been identified regarding data size, head diversity, and label reliability. To address these issues, the 2020 dataset has been reexamined, relabeled, and complemented by adding 1722 images from 5 additional countries, allowing for 81,553 additional wheat heads. We now release in 2021 a new version of the Global Wheat Head Detection dataset, which is bigger, more diverse, and less noisy than the GWHD-2020 version.

Bibliographic Details

Main Authors:	David, E., Serouart, M., Smith, D., Madec, S., Velumani, K., Shouyang Liu, Xu Wang, Pinto Espinosa, F., Shafiee, S., Tahir, I.S.A., Hisashi Tsujimoto, Shuhei Nasuda, Bangyou Zheng, Kirchgessner, N., Aasen, H., Hund, A., Sadhegi-Tehran, P., Koichi Nagasawa, Goro Ishikawa, Dandrifosse, S., Carlier, A., Dumont, B., Mercatoris, B., Evers, B., Ken Kuroki, Haozhou Wang, Masanori Ishii, Badhon, M.A., Pozniak, C.J., LeBauer, D.S., Lillemo, M., Poland, J.A., Chapman, S., Solan, B.de, Baret, F., Stavness, I., Wei Guo
Format:	Article biblioteca
Language:	English
Published:	Science Partner Journals 2021
Subjects:	AGRICULTURAL SCIENCES AND BIOTECHNOLOGY, AGRICULTURAL SCIENCES, DATA, DATA COLLECTION, WHEAT,
Online Access:	https://hdl.handle.net/10883/21706
Tags:	Add Tag No Tags, Be the first to tag this record!