Plants make up more than 80% of the planet’s biomass, and like other living beings, they are part of a diversity that is largely unknown to us. The scientific community has identified some 400,000 species of plants and it is likely that everyone, at some point, has asked themselves: What tree is that in front of me? Applications that use artificial intelligence to identify them from photographs in the field have become popular, bringing the public closer to nature and botany. However, most are not accurate. The ones that work best tend to have very technical terminologies, while the most popular ones frequently fail.
A study published in the journal Plos One He states that some of the popular apps for identifying plants through photos can be as accurate as 4%. The researchers used Google Lens, iNaturalist, Leaf Snap, PlantNet, Plant Snap, and Seek to identify 38 species of herbaceous plants in their natural habitats in Ireland. The conclusion was that most of the apps gave unreliable results, and even the best performing one did not exceed 88% accuracy.
In general, the apps identify flowers better than leaves because of the variety of colors and shapes they have. But iNaturalistfor example, which has a global database and the collaboration of the National Geographic Society, was only able to correctly target 3.6% of the flowers in the study. Plant Snap accurately identified 35.7% of the flowers and 17.1% of the leaves. In total, the one that scored best was Plant Net, with 88.2% success in the flowers. Despite this, it has failed to correctly identify between 12% and 20% of the plants, highlighting that even well-functioning applications need improvement.
Julie Peacock, the study’s researcher, stresses that the “location of the flora can influence the results, since some are better known than others. When apps use machine learning and collect information, the results are likely to be better in the geographic area where the app is most frequently used,” explains the Associate Professor of Ecology at the University of Leeds (UK). Another problem, the study says, is that apps more assertive may be inappropriate for some users due to the wide range of botanical information and terminology that is often unfamiliar to the general public.
And it’s also a question of the identification methodology. Everything counts: the scoring system used, how the app compares the images submitted, and the quality of those photos. “In citizen science there is always a risk that the public will introduce errors into the data. However, it is a great way to get a lot of information. Apps with plant photos verified by expert botanists may have fewer errors, but generating such a large data set would be very difficult,” he adds.
Despite this, the researchers argue that these applications have enormous potential to bring the public closer to the great universe of plants and nature and people should continue using them. But the most important thing is “not just to accept the suggestions, but to consult and see what is the most likely correct answer,” says Peacock. Mainly, if the plant in question can be toxic or harmful.
Among the risks of having a wrong identification, it could be a situation in which a person comes across (better “finds”) a supposedly edible herb and decides to try it, but in reality it is a poisonous one; or if it is identified as an invasive plant, but it is an endangered species. So it is essential to verify the veracity before touching, tearing or eating them. “I encourage people to use them to learn more about the local vegetation,” continues the teacher.
Arbolapp, the Spanish version for trees
The Arbolapp application, created by members of the Higher Council for Scientific Research (CSIC) and the Royal Botanical Garden (RJB), allows the identification of 122 species of wild trees present in Spain, which include native and non-native trees that grow naturally in the territory . Shrubs and species cultivated on agricultural land are excluded from the inventory. Eduardo Actis, coordinator of the project, explains that they have never “intended to cover all the plants in the world”, but to allow the Spaniards to identify any tree that is in front of them. And in a simple way.
The application does not require an internet connection to work, which facilitates its use in isolated natural environments. The identification is done through a guided search, in which the user answers questions about the shape of the leaves, their number or their size. You can also do an open search, by the territory where it is found, by the common or scientific name, or characteristics such as whether it has flowers, fruits or smells. “Without realizing it, you learn to observe the trees, to differentiate their characters and even to use precise terminology to do so,” he continues.
The communication manager believes that image recognition applications are very useful and have the virtue of including a large number of plants, which is why they are complementary tools. “If you want to learn something about Iberian or Canary trees while identifying them, Arbolapp is still the best option,” he concludes.
Since plants do not understand borders, in 2017 the team also inaugurated Arbolapp Canarias, since the flora of the archipelago has little to do with that of the peninsula. Felipe Castilla, a botanical biologist at the RJB and creator of the textual and graphic content of Arbolapp, tells EL PAÍS that each of the species has a “fairly exhaustive” information sheet, which includes ecology, uses, curiosities and anecdotes and if they are protected by law. They also specify if the plant is toxic or poisonous, unlike many of the apps that only present the nomenclature.
Of the almost 400,000 species of plants known to science, some are morphologically very similar or have characteristics that are difficult to recognize in a photo, such as the presence of hairs. “It is clear that the applications will improve, but I doubt that they will ever be perfect, in the same way that even the best botanists make mistakes and know certain floras better than others”, qualifies the expert from the University of Leeds.
With regard to failures when identifying a species, Castilla agrees that it is impossible to reach perfection, even if it is reviewed by botanical experts. For example, there are cases in which plants can belong to one family or another depending on the characteristics. And species evolve. “Nothing is absolute. And less in nature”, concludes the biologist.
