@article{fdi:010092750,
  title = {{E}nhancing plant morphological trait identification in herbarium collections through deep learning-based segmentation},
  author = {{A}riouat, {H}anane and {S}klab, {Y}oucef and {P}rifti, {E}di and {Z}ucker, {J}ean-{D}aniel and {C}henin, {E}ric},
  editor = {},
  language = {{ENG}},
  abstract = {{P}remise{D}eep learning has become increasingly important in the analysis of digitized herbarium collections, which comprise millions of scans that provide valuable resources for studying plant evolution and biodiversity. {H}owever, leveraging deep learning algorithms to analyze these scans presents significant challenges, partly due to the heterogeneous nature of the non-plant material that forms the background of the scans. {W}e hypothesize that removing such backgrounds can improve the performance of these algorithms.{M}ethods{W}e propose a novel method based on deep learning to segment and generate plant masks from herbarium scans and subsequently remove the non-plant backgrounds. {T}he semi-automatic preprocessing stages involve the identification and removal of non-plant elements, substantially reducing the manual effort required to prepare the training dataset.{R}esults{T}he results highlight the importance of effective image segmentation, which achieved an {F}1 score of up to 96.6%. {M}oreover, when used in classification models for plant morphological trait identification, the images resulting from segmentation improved classification accuracy by up to 3% and {F}1 score by up to 7% compared to non-segmented images.{D}iscussion{O}ur approach isolates plant elements in herbarium scans by removing background elements to improve classification tasks. {W}e demonstrate that image segmentation significantly enhances the performance of plant morphological trait identification models.},
  keywords = {deep learning ; herbarium scans ; semantic segmentation ; trait ; classification},
  booktitle = {},
  journal = {{A}pplications in {P}lant {S}ciences},
  volume = {[{E}arly access]},
  numero = {},
  pages = {[12 p.]},
  ISSN = {2168-0450},
  year = {2025},
  DOI = {10.1002/aps3.70000},
  URL = {https://www.documentation.ird.fr/hor/fdi:010092750},
}