FN Archimer Export Format
PT J
TI A new method to control error rates in automated species identification with deep learning algorithms
BT 
AF Villon, Sébastien
   Mouillot, David
   Chaumont, Marc
   Subsol, Gérard
   Claverie, Thomas
   Villéger, Sébastien
AS 1:1,2;2:1,5;3:2,3;4:2;5:1,4;6:1;
FF 1:;2:;3:;4:;5:;6:;
C1 MARBEC, Univ of Montpellier, CNRS, IRD, Ifremer, Montpellier, France
   Research-Team ICAR, LIRMM, Univ of Montpellier, CNRS, Montpellier, France
   University of Nîmes, Nîmes, France
   CUFR Mayotte, Dembeni, France
   Australian Research Council Centre of Excellence for Coral Reef Studies, James Cook University, Townsville, QLD, 4811, Australia
C2 UNIV MONTPELLIER, FRANCE
   UNIV MONTPELLIER, FRANCE
   UNIV NIMES, FRANCE
   UNIV MAYOTTE, FRANCE
   UNIV JAMES COOK, AUSTRALIA
UM MARBEC
IN WOS Cotutelle UMR
   DOAJ
   copubli-france
   copubli-univ-france
   copubli-int-hors-europe
IF 4.379
TC 23
UR https://archimer.ifremer.fr/doc/00640/75244/75406.pdf
   https://archimer.ifremer.fr/doc/00640/75244/75407.docx
LA English
DT Article
AB Processing data from surveys using photos or videos remains a major bottleneck in ecology. Deep Learning Algorithms (DLAs) have been increasingly used to automatically identify organisms on images. However, despite recent advances, it remains difficult to control the error rate of such methods. Here, we proposed a new framework to control the error rate of DLAs. More precisely, for each species, a confidence threshold was automatically computed using a training dataset independent from the one used to train the DLAs. These species-specific thresholds were then used to post-process the outputs of the DLAs, assigning classification scores to each class for a given image including a new class called “unsure”. We applied this framework to a study case identifying 20 fish species from 13,232 underwater images on coral reefs. The overall rate of species misclassification decreased from 22% with the raw DLAs to 2.98% after post-processing using the thresholds defined to minimize the risk of misclassification. This new framework has the potential to unclog the bottleneck of information extraction from massive digital data while ensuring a high level of accuracy in biodiversity assessment. 
PY 2020
PD JUN
SO Scientific Reports
SN 2045-2322
PU Springer Science and Business Media LLC
VL 10
IS 1
UT 000546533700003
DI 10.1038/s41598-020-67573-7
ID 75244
ER

EF