Convolutional neural network application ...
Type de document :
Compte-rendu et recension critique d'ouvrage
Titre :
Convolutional neural network application on a new middle Eocene radiolarian dataset
Auteur(s) :
Carlsson, Veronica [Auteur]
Évolution, Écologie et Paléontologie (Evo-Eco-Paleo) - UMR 8198 [Evo-Eco-Paléo (EEP)]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Danelian, Taniel [Auteur]
Évolution, Écologie et Paléontologie (Evo-Eco-Paleo) - UMR 8198 [Evo-Eco-Paléo (EEP)]
Tetard, Martin [Auteur]
GNS Science
Meunier, Mathias [Auteur]
Évolution, Écologie et Paléontologie (Evo-Eco-Paleo) - UMR 8198 [Evo-Eco-Paléo (EEP)]
Boulet, Pierre [Auteur]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Devienne, Philippe [Auteur]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Ventalon, sandra [Auteur]
Laboratoire d’Océanologie et de Géosciences (LOG) - UMR 8187 [LOG]
Évolution, Écologie et Paléontologie (Evo-Eco-Paleo) - UMR 8198 [Evo-Eco-Paléo (EEP)]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Danelian, Taniel [Auteur]
Évolution, Écologie et Paléontologie (Evo-Eco-Paleo) - UMR 8198 [Evo-Eco-Paléo (EEP)]
Tetard, Martin [Auteur]
GNS Science
Meunier, Mathias [Auteur]
Évolution, Écologie et Paléontologie (Evo-Eco-Paleo) - UMR 8198 [Evo-Eco-Paléo (EEP)]
Boulet, Pierre [Auteur]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Devienne, Philippe [Auteur]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Ventalon, sandra [Auteur]
Laboratoire d’Océanologie et de Géosciences (LOG) - UMR 8187 [LOG]
Titre de la revue :
Marine Micropaleontology
Pagination :
102268
Éditeur :
Elsevier
Date de publication :
2023
ISSN :
0377-8398
Mot(s)-clé(s) en anglais :
middle Eocene
radiolaria
convolutional neural network
image database
automated identification
image recognition
radiolaria
convolutional neural network
image database
automated identification
image recognition
Discipline(s) HAL :
Planète et Univers [physics]/Sciences de la Terre/Paléontologie
Informatique [cs]/Réseau de neurones [cs.NE]
Informatique [cs]/Traitement des images [eess.IV]
Planète et Univers [physics]/Sciences de la Terre/Stratigraphie
Informatique [cs]/Réseau de neurones [cs.NE]
Informatique [cs]/Traitement des images [eess.IV]
Planète et Univers [physics]/Sciences de la Terre/Stratigraphie
Résumé en anglais : [en]
A new radiolarian image database was used to train a Convolutional Neural Network (CNN) for automatic image classification. The focus was on 39 commonly occurring nassellarian species, which are important for biostratigraphy. ...
Lire la suite >A new radiolarian image database was used to train a Convolutional Neural Network (CNN) for automatic image classification. The focus was on 39 commonly occurring nassellarian species, which are important for biostratigraphy. The database consisted of tropical radiolarian assemblages from 129 middle Eocene samples retrieved from ODP Holes 1258A, 1259A, and 1260A (Demerara Rise). A total of 116 taxonomic classes were established, with 96 classes used for training a ResNet50 CNN. To represent the diverse radiolarian assemblage, some classes were formed by grouping forms based on external morphological criteria. This approach resulted in an 86.6% training accuracy. A test set of 800 images from new samples obtained from Hole 1260A was used to validate the CNN, achieving a 75.69% accuracy. The focus then shifted to 39 well-known nassellarian 2 species, using a total of 15 932 images from the new samples. The goal was to determine if the targeted species were correctly classified and explore potential real-world applications of the trained CNN. Different prediction threshold values were experimented with. In most cases, a lower threshold value was preferred to ensure that all species were captured in the correct groups, even if it resulted in lower accuracies within the classes.Lire moins >
Lire la suite >A new radiolarian image database was used to train a Convolutional Neural Network (CNN) for automatic image classification. The focus was on 39 commonly occurring nassellarian species, which are important for biostratigraphy. The database consisted of tropical radiolarian assemblages from 129 middle Eocene samples retrieved from ODP Holes 1258A, 1259A, and 1260A (Demerara Rise). A total of 116 taxonomic classes were established, with 96 classes used for training a ResNet50 CNN. To represent the diverse radiolarian assemblage, some classes were formed by grouping forms based on external morphological criteria. This approach resulted in an 86.6% training accuracy. A test set of 800 images from new samples obtained from Hole 1260A was used to validate the CNN, achieving a 75.69% accuracy. The focus then shifted to 39 well-known nassellarian 2 species, using a total of 15 932 images from the new samples. The goal was to determine if the targeted species were correctly classified and explore potential real-world applications of the trained CNN. Different prediction threshold values were experimented with. In most cases, a lower threshold value was preferred to ensure that all species were captured in the correct groups, even if it resulted in lower accuracies within the classes.Lire moins >
Langue :
Anglais
Vulgarisation :
Non
Projet ANR :
Source :
Fichiers
- document
- Accès libre
- Accéder au document
- accepted_version_to%20HAL_%28clean%29VC_revised_MS.pdf
- Accès libre
- Accéder au document