@article{fdi:010082629,
  title = {{L}everaging social media and deep learning to detect rare megafauna in video surveys},
  author = {{M}annocci, {L}. and {V}illon, {S}. and {C}haumont, {M}. and {G}uellati, {N}. and {M}ouquet, {N}. and {I}ovan, {C}orina and {V}igliola, {L}aurent and {M}ouillot, {D}.},
  editor = {},
  language = {{ENG}},
  abstract = {{D}eep learning has become a key tool for the automated monitoring of animal populations with video surveys. {H}owever, obtaining large numbers of images to train such models is a major challenge for rare and elusive species because field video surveys provide few sightings. {W}e designed a method that takes advantage of videos accumulated on social media for training deep-learning models to detect rare megafauna species in the field. {W}e trained convolutional neural networks ({CNN}s) with social media images and tested them on images collected from field surveys. {W}e applied our method to aerial video surveys of dugongs ({D}ugong dugon) in {N}ew {C}aledonia (southwestern {P}acific). {CNN}s trained with 1303 social media images yielded 25% false positives and 38% false negatives when tested on independent field video surveys. {I}ncorporating a small number of images from {N}ew {C}aledonia (equivalent to 12% of social media images) in the training data set resulted in a nearly 50% decrease in false negatives. {O}ur results highlight how and the extent to which images collected on social media can offer a solid basis for training deep-learning models for rare megafauna detection and that the incorporation of a few images from the study site further boosts detection accuracy. {O}ur method provides a new generation of deep-learning models that can be used to rapidly and accurately process field video surveys for the monitoring of rare megafauna.},
  keywords = {convolutional neural networks ; endangered megafauna ; internet ecology ; monitoring ; species detection ; deteccion de especies ; ecologia de internet ; megafauna en peligro ; monitoreo ; redes neurales ; convolucionales ; {NOUVELLE} {CALEDONIE}},
  booktitle = {},
  journal = {{C}onservation {B}iology},
  volume = {36},
  numero = {1},
  pages = {e13798[11 ]},
  ISSN = {0888-8892},
  year = {2022},
  DOI = {10.1111/cobi.13798},
  URL = {https://www.documentation.ird.fr/hor/fdi:010082629},
}