Gast ::
| Hauptseite > Publikationsdatenbank > Growliflower: An image time series dataset for growth analysis of cauliflower |
| Hauptseite > Publikationsdatenbank > Growliflower: An image time series dataset for growth analysis of cauliflower |
| Conference Presentation (After Call) | FZJ-2023-00134 |
; ; ; ; ; ; ; ;
2022
Abstract: In our video, we present our benchmark dataset GrowliFlower. It contains weekly, georeferenced UAV captured image time series of two fields sized 0.39 to 0.60 ha for one growing period of cauliflower in 2020 and 2021 each. We extract and provide image time series for thousands of individual plants and collect in-situ reference data in the field. The reference data contain phenotypic traits such as phenological development, diameter, height, head size and more. Additionally, we have defoliated cauliflower heads and capture image data before and after defoliation. Furthermore, we provide pixel-precise leaf and plant instance segmentation as well as stem annotations. Our benchmark is used to develop and evaluate machine learning models for instance for classification, detection, semantic segmentation, instance segmentation or stem detection tasks, but also for time series analysis. An example for the application of our benchmark can be the analysis of plant development and the determination of the harvest time of cauliflower. The entire dataset will be published and publicly accessible.
|
The record appears in these collections: |