@article{fdi:010095889,
  title = {{P}lant{SAM} : an object detection-driven segmentation pipeline for herbarium specimens},
  author = {{S}klab, {Y}oucef and {C}astanet, {F}. and {A}riouat, {H}. and {A}rib, {S}. and {Z}ucker, {J}ean-{D}aniel and {C}henin, {E}ric and {P}rifti, {E}di},
  editor = {},
  language = {{ENG}},
  abstract = {{P}remise {D}eep learning-based classification of herbarium images is hampered by background heterogeneity, which introduces noise and artifacts that can potentially mislead models and degrade their accuracy. {A}ddressing these effects is essential to enhance overall performance. {M}ethods {W}e introduce {P}lant{SAM}, an automated segmentation pipeline that integrates {YOLO}v10 for plant region detection and the {S}egment {A}nything {M}odel ({SAM}2) for segmentation. {YOLO}v10 generates bounding box prompts to guide {SAM}2, enhancing segmentation accuracy. {B}oth models were fine-tuned on herbarium images and evaluated using intersection over union ({I}o{U}) and {S} & oslash;rensen-{D}ice coefficient metrics.{R}esults {P}lant{SAM} achieved state-of-the-art segmentation performance, with an {I}o{U} of 0.94 and a {S} & oslash;rensen-{D}ice coefficient of 0.97. {I}ncorporating segmented images into classification models led to consistent performance improvements across five tested botanical traits, with accuracy gains of up to 4.36% and {F}1 score improvements of 4.15%.{C}onclusions {O}ur findings highlight the importance of background removal in herbarium image analysis, as it significantly enhances classification performance by enabling models to focus more effectively on the foreground plant structures.},
  keywords = {botanical analysis ; herbarium specimens ; {S}egment {A}nything {M}odel ({SAM}) ; semantic segmentation ; {YOLO}v10},
  booktitle = {},
  journal = {{A}pplications in {P}lant {S}ciences},
  volume = {[{E}arly access]},
  numero = {},
  pages = {[14 p.]},
  ISSN = {2168-0450},
  year = {2025},
  DOI = {10.1002/aps3.70034},
  URL = {https://www.documentation.ird.fr/hor/fdi:010095889},
}