Abstract
Leaf counting in potted plants is an important building block for estimating their health status and growth rate and has obtained increasing attention from the visual phenotyping community in recent years. Two novel deep learning approaches for visual leaf counting tasks are proposed, evaluated, and compared in this study. The first method performs counting via direct regression but using multiple image representation resolutions to attend leaves of multiple scales. The leaf count from multiple resolutions is fused using a novel technique to get the final count. The second method is detection with a regression model that counts the leaves after locating leaf center points and aggregating them. The algorithms are evaluated on the Leaf Counting Challenge (LCC) dataset of the Computer Vision Problems in Plant Phenotyping (CVPPP) conference 2017, and a new larger dataset of banana leaves. Experimental results show that both methods outperform previous CVPPP LCC challenge winners, based on the challenge evaluation metrics, and place this study as the state of the art in leaf counting. The detection with regression method is found to be preferable for larger datasets when the center-dot annotation is available, and it also enables leaf center localization with a 0.94 average precision. When such annotations are not available, the multiple scale regression model is a good option.
Highlights
Object counting is important for a variety of tasks in the agriculture and phenotyping domains
There is no clear winner in this examination, but the proposed Multiple-Scale Regression (MSR) and Detection with Regression Network (DRN) have a slight advantage over the Leaf Counting Challenge (LCC) winner (Dobrescu et al, 2017)
It can be seen that when the training set is small, the MSR method has an advantage, while for larger datasets (A4 and Banana leaves (BL)) the DRN method usually performs better
Summary
Object counting is important for a variety of tasks in the agriculture and phenotyping domains. With the growing need for systematic plant phenotyping (Großkinsky et al, 2015) and the development of recent Convolutional Neural Network (CNN)-based techniques (Ren et al, 2015; He et al, 2016), visual leaf counting has attracted considerable attention (Giuffrida et al, 2016; Dobrescu et al, 2017; Lu et al, 2017; Teimouri et al, 2018; Jiang et al, 2019; Kuznichov et al, 2019). The resulting network provides the state-of-the-art results in the Leaf Segmentation Challenge (LSC), which is a segmentation plus counting benchmark on the same datasets (Scharr et al, 2014; Bell and Dee, 2016; Minervini et al, 2016) as the LCC Such segmentation-based methods have a significant drawback for counting since only successfully segmented leaves are counted, and their results in the counting evaluation metric are less competitive
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
