@TechReport{TR155,
  author      = "Fabian Wolf",
  title       = "Learning-based Leaf Cell Instance Segmentation",
  institution = "PRIP, TU Wien",
  number      = "PRIP-TR-155",
  year        = "2022",
  url         = "https://www.prip.tuwien.ac.at/pripfiles/trs/tr155.pdf",
  abstract    = "Due to climate change, variability of temperature and rainfall increases. Based on this, understanding changes in leaf anatomy and function under those conditions is desirable. We aim to advance the understanding of changes in leaf anatomy by investigating deep- learning based instance segmentation of leaf cells. Accordingly, state-of-the-art (cell) in- stance segmentation methods are discussed. Based on the discussion, we investigated 2D- and 3D-Cellpose with and without additional semantic information. In general, we found that Cellpose is not suited well for the task, but might be suited for segmenting roundish (sponge and palisade) cells. We conclude that further research is required and propose to investigate embedding-based approaches.",
}