Deep Green

Project Details


Österreichische Bundesforste AG


2009/11/01 – 2010/06/31


Robert Sablatnig


Stefan Fiel

Correct identification of tree species from leaves, needles and bark is a task which requires expertise. With increasing capabilities of mobile devices, like resolution of the integrated cameras and more computing power these devices can be used to take over this task. Within a project with the “Österreichischen Bundesforste AG” (“Austrian Federal Forests”) the main goal is to automatically classify tree species from images of the leaves, needles or bark. Children, students and interested adults for instance can identify the plant species during a walk with their mobile devices for pedagogic reasons.

The classification should work with the 12 most common austrian trees:

  • Abele (Weißpappel)
  • Ash (Esche)
  • Beech (Buche)
  • Black pine (Schwarzkiefer)
  • Fir (Fichte)
  • Hornbeam (Hainbuche)
  • Larch (Lärche)
  • Mountain oak (Traubeneiche)
  • Scotch pine (Rotkiefer)
  • Spruce (Tanne)
  • Swiss stone pine (Zirbe)
  • Sycamore maple (Bergahorn)

Within an experts meeting a competition between Deep Green, a forester and a biologist has been carried out. The goal of this test is to show wheter the information in the images are sufficient for a correct classification or if experts use additional information, like buds, the habit, site, and the surrounding of the tree. See:

Revierleiter schlägt Computer [DE](Österreichische Bauernzeitung)


S. Fiel and R. Sablatnig, “Leaf classification using local features”, In 34th Annual Workshop of the Austrian Association f. Pattern Recognition (ÖAGM 2010), Zwettl, Austria, pp. 123-130, May 2010.