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.

Download full-text PDF

Source
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC8548052PMC
http://dx.doi.org/10.34133/2021/9846158DOI Listing

Publication Analysis

Top Keywords

wheat head
16
head detection
16
global wheat
12
wheat heads
8
head
5
wheat
5
detection
4
detection 2021
4
2021 improved
4
dataset
4

Similar Publications

Want AI Summaries of new PubMed Abstracts delivered to your In-box?

Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!