Close
< Return

Identification aids

Who hasn't experienced this: sometimes it's hard to identify a plant. Then you are glad to get some help, to have a look at another identification key or to get a second opinion. Or sometimes the detail to be observed is simply too small for the naked eye and a magnifying aid is needed. On this page you will find an explanation of the identification aids in FlorApp, a selection of helpful floras and websites, as well as a comparison of different magnification aids.

 

 

Identification aids in FlorApp

Since version 3.0, an identification tool has been integrated directly into FlorApp. If a photo is uploaded and you're connected to the internet, FlorID can be consulted. But even without a photo and internet connection, there is a little help: if you already have an initial suspicion about the identity of a species, you can display a list of species that are often confused with the suspected species.

 

 

FlorID

FlorID is a free plant identification tool for Swiss species. The tool is the result of the COMECO project, a collaboration between InfoFlora and the WSL (Swiss Federal Institute for Forest, Snow and Landscape Research). The tool identifies plants from uoploaded photos. In FlorApp, the tool also takes into account the location and the date of the observation. This way, the probability (by ecology and known distribution) that a certain species occurs at that location and is in that (phenological) state at that date is taken into account. The result of the analysis is a list of species, each with an associated percentage value. This value represents a confidence value. As a rule of thumb, a low confidence value indicates an uncertain identification, whereas a high confidence value is no guarantee for a correct identification.

Image UploadResults

FlorID is based on artificial intelligence. It is a so-called "classifier", which is an algorithm (a manual for computers) that classifies data into predefined categories. In the case of FlorID, the data are photos and the categories are plant species. For the classification, the algorithm takes into account different layers of information. FlorID uses the entered images, the location and the date of the observation as layers.

Before it can be applied, a classifier has to be "trained". By means of the pre-classified training data entered, the algorithm "learns" how to classify future input data. The FlorID classifier was trained with millions of quality-checked images from Citizen Scientists and Experts. In addition, the distribution of species was modelled using even more geolocated observations. This model in turn was also used for the training of FlorID. You can think of it as a small child learning through experience to categorise dogs and fish. And that if something flies, it is probably a bird rather than a fish. (Since these kinds of classifiers are modelled on the human brain, they are also called "artificial neural networks").

When you now enter photos, FlorID compares the new image material with the photos used as training data. It also takes into account the probability that a species occurs at a certain location and is in a certain (phenological) state at a certain time of year. These three layers of information are then calculated into a confidence score and the species with the highest confidence scores are listed as the result.

The focus of FlorID is on naturally occurring species. With over 2500 taxa, it can distinguish a large proportion of the wild plant species of the Swiss flora, including a large list of alien species that are commonly found in nature. However, the tool is not designed to identify ornamental plants that do not disperse in the wild. Thus, no correct identification can be expected for cultivated or garden plants.

Currently, plants are identified at the species or species aggregate level. Subspecies are therefore not yet taken into account.

In contrast to other image recognition tools, FlorID takes into account not only the images entered but also the location and date of the observation. Furthermore, the accuracy also benefits from being limited to the Swiss wild flora.

FlorID has only been trained to detect species that occur in the wild in Switzerland. It can therefore only recognise these species. If a photo of another species is entered, the Swiss species most similar to it is presented as the most probable.

It is also essential to understand that FlorID does not compare individual features, but only the entirety of the pixels of an image. Thus, it does not see inferior ovaries or opposite leaves and therefore cannot systematically restrict to a family or genus as we humans are able to.

  • Always question the result critically: does it make sense? is it probable?
  • A low confidence value indicates an uncertain identification but a high confidence value is no guarantee for a correct identification.
  • More and better photos may improve the result

Literature

Eine (unvollständige) Auswahl von Floren, die weiterhelfen können:

Schweiz:

  • Flora Helvetica, Illustrierte Flora der Schweiz, 6. Aufl. (Lauber et al. 2018) 
  • Flora Helvetica, Exkursionsflora, 2.Aufl. (Eggenberg et al. 2022) 
  • Der "Binz": Schul- und Exkursionsflora für die Schweiz (Lenzin & Heitz-Weniger 2022); Le nouveau Binz (Aeschimann & Burdet 2003)
  • Flora der Schweiz (Hess et al. 1980)

Nachbarländer, Alpen:

  • Flora Gallica (Tison & Foucault 2014)
  • Rothmaler, Exkursionsflora von Deutschland, 22. Aufl. (Müller et al. (Hrsg.) 2021) (Grundband, Atlasband + Kritischer Ergänzungsband)
  • Flora Germanica (Hassler & Muer 2022)
  • Exkursionsflora für Österreich, Liechtenstein und Südtirol. 3. Aufl. (Fischer et al. 2008)
  • Flora d'Italia (Pignatti 1982 (wegen Illustrationen und Karten weiterhin relevant), 2. Aufl. Pignatti et al. 2017-2019)
  • Flora Alpina (Aeschimann et al. 2004)
  • Online-Exkursionsflora der Alpen und angrenzender Gebiete (Götz 2021)

Weiteres Europa:

  • Nouvelle Flore de la Belgique, du Grand-Duché de Luxembourg, du nord de la France et des régions voisines (Lambinon et al. 2003)
  • Flora Iberica (auch online verfügbar)
  • Gustav Hegi, Illustrierte Flora von Mitteleuropa, 3. Aufl. (Conert et al. (Hrsg.) 1979-1992)
  • Den Nya Nordiska Floran (Mossberg & Stenberg 2003)

Zierpflanzen und Pflanzen mit Einbürgerungspotenzial:

  • Rothmaler Band 5: Krautige Zier- und Nutzpflanzen (Jäger et a. (Hrsg.) 2007)
  • Flora der Gehölze, 5. Aufl. (Roloff & Bärtels 2018)
  • New Flora of the British Isles, 4th ed. (Stace 2019)
  • Flora of North America (Online-Version)

 

 

Online Resources

Eine (unvollständige) Auswahl von Webseiten, die weiterhelfen können:

Lupe, Binokular, Mikroskop & Co.

Wie rüstet man sich aus, um Details zu sehen, die zu klein sind, um mit dem blossen Auge beobachtet zu werden? Zu welchem Preis und mit welchem Ergebnis? Im FloraCH-Artikel vom Herbst 2022 wurden die Ergebnisse mehrerer Vergrösserungshilfen anhand von drei konkreten Beispielen präsentiert. Hier finden Sie einige nützliche Ergänzungen zu den getesteten Geräten und zur Focus Stacking-Technik.

FloraCH-Artikel "Die Grenzen des Sichtbaren erweitern"

Ergänzungen zum Artikel