@article{fdi:010083828,
  title = {{A}dvanced image recognition : a fully automated, high-accuracy photo-identification matching system for humpback whales},
  author = {{C}heeseman, {T}. and {S}outherland, {K}. and {P}ark, {J}. and {O}lio, {M}. and {F}lynn, {K}. and {C}alambokidis, {J}. and {J}ones, {L}. and {G}arrigue, {C}laire and {J}ordan, {A}. {F}. and {H}oward, {A}. and {R}eade, {W}. and {N}eilson, {J}. and {G}abriele, {C}. and {C}lapham, {P}.},
  editor = {},
  language = {{ENG}},
  abstract = {{W}e describe the development and application of a new convolutional neural network-based photo-identification algorithm for individual humpback whales ({M}egaptera novaeangliae). {T}he method uses a {D}ensely {C}onnected {C}onvolutional {N}etwork ({D}ense{N}et) to extract special keypoints of an image of the ventral surface of the fluke and then a separate {D}ense{N}et trained to look for features within these keypoints. {T}he extracted features are then compared against those of the reference set of previously known humpback whales for similarity. {T}his offers the potential to successfully automate recognition of individuals in large photographic datasets such as in ocean basin-wide marine mammal studies. {T}he algorithm requires minimal image pre-processing and is capable of accurate, rapid matching of fair to high-quality humpback fluke photographs. {I}n real world testing compared to manual image matching, the algorithm reduces image management time by at least 98% and reduces error rates of missing potential matches from approximately 6-9% to 1-3%. {T}he success of this new system permits automated comparisons to be made for the first time across photo-identification datasets with tens to hundreds of thousands of individually identified encounters, with profound implications for long-term and large population studies of the species.},
  keywords = {{A}utomated image recognition ; {C}omputer vision ; {D}eep convolutional neural networks ; {K}aggle competition ; {M}achine learning ; {M}ark recapture ; {M}egaptera novaeangliae ; {P}hoto-{ID}},
  booktitle = {},
  journal = {{M}ammalian {B}iology},
  volume = {102},
  numero = {3},
  pages = {[15 p.]},
  ISSN = {1616-5047},
  year = {2022},
  DOI = {10.1007/s42991-021-00180-9},
  URL = {https://www.documentation.ird.fr/hor/fdi:010083828},
}