Fig 1.
Coverage of the area of the Czech Republic by the records in the database.
The map shows the number of observation records from the given area, with the legend in logarithmic scale.
Fig 2.
Method used to scan a living specimen with a desktop scanner.
Fig 3.
Sample screens from the application.
a) interface of the classification algorithm, b) catalogue of odonates, c) information about the species.
Fig 4.
Histogram of the positions of correct classifications in a list sorted by relevance (evaluated on the testing dataset).
Fig 5.
Operation of the mobile application.
The application’s classification model is based on historical and recent presence data and expert knowledge. The amateur user inputs observed dragonfly description into the application and obtains a list of probable species. The user then selects the most appropriate classification from the list (possibly also by comparing the observed species with the photograph) and sends the observation report to the central database, where the expert biologist verifies the records and the amateur user is notified. Stored presence records will be used to update the classification model after each year in the future.