The ability to identify plant species in the field is very important to the study of plants and it is also of interest to a more general audience including nature lovers, gardeners, and hikers. There has been a recent boom in the availability of smartphone apps that aid in the identification of plants in the field. Many of these apps rely on artificial intelligence, in which a computer algorithm is used to compare a user captured image to an online database. Increasing the size of the database of images which can be used for identification has been one key to the rapid recent advances in the field. Image recognition technology has developed even further in the past few years, with some apps now using deep learning algorithms for identification, yet it remains unclear whether smartphone apps have the consistency and accuracy required for use in field botany studies.
In his new viewpoint article published in AoBP, Hamlyn Jones reviews and compares several free smartphone apps that attempt to automatically identify unknown plants from images taken in natural environments. Using images of plants growing wild in Britain, Jones tested nine free apps or websites, including PlantSnap and Google Lens. Plant identification apps are continually improving but Jones’ results suggest that the best ones (in this case Plant.id and Flora Incognita) already have an outstanding success rate. These apps were able to identify the correct species in roughly half of the plant images used in the study and identified to the correct plant family in up to three quarters of images. Although this accuracy is relatively high, Jones suggests that, for any quantitative biodiversity study or for ecological surveys, there remains a need for validation by experts or against conventional floras, particularly for rare or hard-to-distinguish species. He concludes by discussing that the selection of which app to use may ultimately depend on the situation, for instance the needs of a field botanist may differ from the needs of an enthusiast. Regardless of which app you choose, he feels the future is bright for plant identification apps and states that their “performance is expected to improve rapidly, especially with the incorporation of crowd-sourced data”.
Following a PhD on plant environmental physiology at the Australian National University, Hamlyn Jones’s research career has been as a plant physiologist particularly concentrating on stress tolerance and environmental physics, and more recently remote sensing, and their application to crop improvement. Among his many publications are two widely used texts (Plants and Microclimate (1983, 1992 & 2014) and Remote sensing of vegetation: principles, techniques and applications (2010)).Throughout this career Hamlyn retained an interest in natural history and plant identification, so on retirement from the University of Dundee this led to his development of a novel internet/smartphone key for the visual identification of British plants (https://visual-flora.org.uk) and a wider interest in the potential of artificial intelligence approaches to plant identification, especially in their use for rigorous studies of biodiversity.