Image recognition using convolutional neural networks for classification of honey bee subspecies - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Apidologie Année : 2022

Image recognition using convolutional neural networks for classification of honey bee subspecies

Dario de Nart
  • Fonction : Auteur
Cecilia Costa
Gennaro Di Prisco
  • Fonction : Auteur
Emanuele Carpana
  • Fonction : Auteur

Résumé

AbstractFour models based on convolutional neural networks were used to investigate whether image recognition techniques applied to honey bee wings could be used to discriminate among honey bee subspecies. A dataset consisting of 9887 wing images belonging to 7 subspecies and one hybrid was analysed with ResNet 50, MobileNet V2, Inception Net V3, and Inception ResNet V2. Accuracy values of classification of individual wings were over 0.92, and all models outperformed traditional morphometric evaluation. The Inception models achieved the highest accuracies and higher scores of precision and recall for most classes. When wing images were grouped by colony, almost all wings in the colony samples were labelled with the same class. We conclude that automatic image recognition and machine learning applied to honey bee wings can reliably discriminate among the European subspecies and could thus represent a useful tool for fast classification of honey bee subspecies for breeding and conservation aims.
Fichier principal
Vignette du fichier
13592_2022_Article_918.pdf (2.62 Mo) Télécharger le fichier
Origine : Publication financée par une institution

Dates et versions

hal-04025698 , version 1 (13-03-2023)

Identifiants

Citer

Dario de Nart, Cecilia Costa, Gennaro Di Prisco, Emanuele Carpana. Image recognition using convolutional neural networks for classification of honey bee subspecies. Apidologie, 2022, 53 (1), pp.5. ⟨10.1007/s13592-022-00918-5⟩. ⟨hal-04025698⟩
14 Consultations
13 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More