LabelStoma: A tool for stomata detection based on the YOLO algorithm

Abstract

Stomata are pores in the epidermal tissue of leaf plants formed by specialised cells called guard cells, which regulate the stomatal opening. Stomata facilitate gas exchange, being pivotal in the regulation of processes such as photosynthesis and transpiration. The analysis of the number and behaviour of stomata is a task carried out by studying microscopic images, and that can serve, among other things, to better manage crops in agriculture or to better understand how plants fix CO₂ and lose water under different conditions. However, quantifying the number of stomata in an image traditionally has been a labor intensive and thus expensive process since an image might contain dozens of stomata. Several automatic stomata detection models have been developed and presented in the literature, but they fail to generalise to images from species different to those employed to train the model; and, in addition, they lack a simple interface to employ them. In this work, we tackle these problems by training a YOLO model. Such a model achieves a F1-score of 0.91 in images from the species employed for training it, and similar F1-score for datasets containing images of different species. Moreover, in order to facilitate the use of the model, we have developed LabelStoma, an open-source and simple-to-use graphical user interface that employs the YOLO model. In addition, this tool provides a simple method to adapt the YOLO model to the users’ images, and, therefore, customising the model to the users’ needs. Thanks to this work, the analysis of plant stomata of different species will be more reliable and comparable; and, the developed tools will help to advance our understanding of CO₂ and H₂O dynamics in plants, such as photosynthesis and transpiration, and ecosystems related processes, such as carbon and water cycles.

Plant phenotyping relevance

気孔を顕微鏡画像から自動検出・定量するYOLOモデルとGUIツールを開発し、異種間性能を検証しており、植物表現型取得法が中心である。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.