Greenotyper: Image-Based Plant Phenotyping Using Distributed Computing and Deep Learning

Abstract

Image-based phenotype data with high temporal resolution offers advantages over end-point measurements in plant quantitative genetics experiments, because growth dynamics can be assessed and analysed for genotype-phenotype association. Recently, network-based camera systems have been deployed as customizable, low-cost phenotyping solutions. Here, we implemented a large, automated image-capture system based on distributed computing using 180 networked Raspberry Pi units that could simultaneously monitor 1,800 white clover ( Trifolium repens ) plants. The camera system proved stable with an average uptime of 96% across all 180 cameras. For analysis of the captured images, we developed the Greenotyper image analysis pipeline. It detected the location of the plants with a bounding box accuracy of 97.98%, and the U-net-based plant segmentation had an intersection over union accuracy of 0.84 and a pixel accuracy of 0.95. We used Greenotyper to analyze a total of 355,027 images, which required 24–36 h. Automated phenotyping using a large number of static cameras and plants thus proved a cost-effective alternative to systems relying on conveyor belts or mobile cameras.

Broad‐scale applications of the Raspberry Pi: A review and guide for biologists

“…A Raspberry Pi can be left to run continuously 24/7, just always make sure that you use a good quality power supply that provides enough amperage. A shortcoming of the Raspberry Pi is that it does not have a realtime clock (RTC) and only updates the time when connected to the F I G U R E 6 Various custom-build Graphical User Interfaces (GUIs) for controlling the Raspberry Pi in an easy-to-use, standardised way, from left to right: PiVR (Tadres & Louis, 2020); pirecorder (Jolles, 2020); FlyPi (Maia Chagas et al, 2017) and Greenotyper (Tausen et al, 2020) internet. A RTC module is therefore recommended for projects without an internet connection (for tutorial, see Table S2).…”

“…Gurdita et al., 2016; Lendvai et al., 2015; McBride & Courter, 2019; Singh et al., 2019). The large number of interfaces and broad connectivity of the Raspberry Pi enable the development of solutions that provide a highly affordable alternative to expensive research equipment that many researchers do not have the budget for (Dolgin, 2018), such as operant conditioning devices, plant phenotyping systems and confocal microscopes (Maia Chagas et al., 2017; Stanton et al., 2020; Tausen et al., 2020). Its low cost also facilitates many devices to be employed simultaneously and enables researchers to try out new ideas, opening the door for creative and novel solutions.…”

“…Tausen et al. (2020) extended the use of these kinds of phenotyping solutions and implemented a large‐scale, automated image‐capture system with a total of 180 networked Raspberry Pi's with which they monitored 1,800 individual plants simultaneously over multiple months. Finally, Gaggion et al.…”

“…, including accurate detection of environmental temperature, humidity, barometric pressure and light levels. Especially in combination with automated messaging services, such as to notify experimenters when variables go out of range (e.g Grindstaff et al, 2019;Gurdita et al, 2016;Tausen et al, 2020),. this helps minimise fluctuations in environmental factors that may influence experiments or the behaviour of animals, and thereby improves the accuracy and reproducibility of experiments.…”

Competition, Nodule Occupancy, and Persistence of Inoculant Strains: Key Factors in the Rhizobium-Legume Symbioses

“…Indeed, more and more multicomponent interaction studies are being conducted ( Vuong et al, 2017 ; Burghardt et al, 2018 ; Gunnabo et al, 2019 ; Batstone et al, 2020 ; Mendoza-Suárez et al, 2020 ; Fagorzi et al, 2021 ), and they are already generating important knowledge to help understand some of these interactions and give more weight to the performance of rhizobial inoculants. Additionally, the rapid progress in NGS with open-source laboratory equipment automation ( Figure 3C ; Wong et al, 2018 ; Faiña et al, 2020 ), and the application of machine learning in big data analysis of microbiome studies ( Cammarota et al, 2020 ; Ghannam and Techtmann, 2021 ) and in biological-image analysis ( Berg et al, 2019 ; Chung et al, 2020 ; Tausen et al, 2020 ), will make multicomponent interactions studies more achievable, providing the opportunity to identify a larger number of elite strains in less time. The introduction of a wider number of variables in the experimental assays conducted to identify elite strains will increase the probability of designing rhizobial inoculants with better performance under field conditions.…”