Sistema de Clasificación Automático de Peces Endémicos del Ecuador Usando Redes Neuronales Convolucionales

Anthony Sánchez-Guashpa, Pablo Pico-Valencia, Pedro Jiménez, Juan A. Holgado-Terriza

Producción científica: Contribución a una revistaArtículorevisión exhaustiva

1 Cita (Scopus)

Resumen

Artificial neural networks are widely used in pattern recognition tasks, including the classification of people, animal species and objects. This paper presents an automatic classifier of fish species endemic to Ecuador using convolutional neural networks. The proposed classifier, developed in Python, consisted of 4 convolutional layers to which 32, 64, 128 and 256 filters were applied respectively for the extraction of features from the input images to the network. This model was trained with a total of 36,140 images of fish species categorized into 52 genera corresponding to 94 fish species. After the training and evaluation of the developed model, the result was an automatic classifier of endemic fishes of Ecuador with an accuracy of 97.55%.

Título traducido de la contribuciónAutomatic Classification System for Endemic Fishes of Ecuador Using Convolutional Neural Networks
Idioma originalEspañol
Páginas (desde-hasta)444-457
Número de páginas14
PublicaciónRISTI - Revista Iberica de Sistemas e Tecnologias de Informacao
Volumen2021
N.ºE45
EstadoPublicada - 2021

Nota bibliográfica

Publisher Copyright:
© 2021, Associacao Iberica de Sistemas e Tecnologias de Informacao. All rights reserved.

Palabras clave

  • CNN
  • Deep learning
  • Ecuador
  • fish
  • Python

Citar esto